wangbo/ComfyUI
1
0
Fork 0
mirror of https://github.com/comfyanonymous/ComfyUI.git synced 2026-06-17 21:39:45 +08:00
Code Issues Actions 14 Packages Projects Releases Wiki Activity

Merge branch 'master' into alexis/add_output_save_nodes

Browse Source
This commit is contained in:
Alexis Rolland 2026-06-17 08:34:44 +08:00 committed by GitHub
commit 7ba2a90f41
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
58 changed files with 591 additions and 258 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
comfy/sd.py
View File

@ -67,6 +67,7 @@ import comfy.text_encoders.anima
import comfy.text_encoders.ace15
import comfy.text_encoders.longcat_image
import comfy.text_encoders.qwen35
import comfy.text_encoders.qwen3vl
import comfy.text_encoders.ernie
import comfy.text_encoders.gemma4
import comfy.text_encoders.cogvideo
@ -1353,6 +1354,8 @@ class TEModel(Enum):
GEMMA_4_31B = 31
T5_GEMMA = 32
GPT_OSS_20B = 33
QWEN3VL_4B = 34
QWEN3VL_8B = 35
def detect_te_model(sd):
@ -1414,6 +1417,8 @@ def detect_te_model(sd):
if weight.shape[0] == 5120:
return TEModel.QWEN35_27B
return TEModel.QWEN35_2B
if "model.visual.deepstack_merger_list.0.norm.weight" in sd: # DeepStack is unique to Qwen3-VL
return TEModel.QWEN3VL_4B if sd["model.visual.merger.linear_fc2.weight"].shape[0] == 2560 else TEModel.QWEN3VL_8B
if "model.layers.0.post_attention_layernorm.weight" in sd:
weight = sd['model.layers.0.post_attention_layernorm.weight']
if 'model.layers.0.self_attn.q_norm.weight' in sd:
@ -1612,6 +1617,16 @@ def load_text_encoder_state_dicts(state_dicts=[], embedding_directory=None, clip
qwen35_type = {TEModel.QWEN35_08B: "qwen35_08b", TEModel.QWEN35_2B: "qwen35_2b", TEModel.QWEN35_4B: "qwen35_4b", TEModel.QWEN35_9B: "qwen35_9b", TEModel.QWEN35_27B: "qwen35_27b"}[te_model]
clip_target.clip = comfy.text_encoders.qwen35.te(**llama_detect(clip_data), model_type=qwen35_type)
clip_target.tokenizer = comfy.text_encoders.qwen35.tokenizer(model_type=qwen35_type)
elif te_model in (TEModel.QWEN3VL_4B, TEModel.QWEN3VL_8B):
if clip_type == CLIPType.IDEOGRAM4 and te_model == TEModel.QWEN3VL_8B: # Ideogram4 reuses the full Qwen3-VL-8B (13-layer tap for conditioning + multimodal generate).
clip_data[0] = comfy.utils.state_dict_prefix_replace(clip_data[0], {"model.language_model.": "model.", "model.visual.": "visual.", "lm_head.": "model.lm_head."})
clip_target.clip = comfy.text_encoders.ideogram4.te_qwen3vl(**llama_detect(clip_data))
clip_target.tokenizer = comfy.text_encoders.ideogram4.Ideogram4Qwen3VLTokenizer
else:
clip_data[0] = comfy.utils.state_dict_prefix_replace(clip_data[0], {"model.language_model.": "model.", "model.visual.": "visual.", "lm_head.": "model.lm_head."})
qwen3vl_type = {TEModel.QWEN3VL_4B: "qwen3vl_4b", TEModel.QWEN3VL_8B: "qwen3vl_8b"}[te_model]
clip_target.clip = comfy.text_encoders.qwen3vl.te(**llama_detect(clip_data), model_type=qwen3vl_type)
clip_target.tokenizer = comfy.text_encoders.qwen3vl.tokenizer(model_type=qwen3vl_type)
elif te_model == TEModel.QWEN3_06B:
clip_target.clip = comfy.text_encoders.anima.te(**llama_detect(clip_data))
clip_target.tokenizer = comfy.text_encoders.anima.AnimaTokenizer

41
comfy/text_encoders/ideogram4.py
View File

@ -9,6 +9,7 @@ import os
from transformers import Qwen2Tokenizer
import comfy.text_encoders.llama
import comfy.text_encoders.qwen3vl
from comfy import sd1_clip
# Reference taps outputs of layers (0,3,...,35); comfy captures layer inputs, offset by +1.
@ -77,3 +78,43 @@ def te(dtype_llama=None, llama_quantization_metadata=None):
model_options["quantization_metadata"] = llama_quantization_metadata
super().__init__(device=device, dtype=dtype, model_options=model_options)
return Ideogram4TEModel_
# Full Qwen3-VL-8B variant with vision
class Ideogram4Qwen3VLClipModel(comfy.text_encoders.qwen3vl.Qwen3VLClipModel):
def __init__(self, device="cpu", dtype=None, attention_mask=True, model_options={}):
super().__init__(device=device, layer=IDEOGRAM4_TAP_LAYERS, layer_idx=None, dtype=dtype,
attention_mask=attention_mask, model_options=model_options, model_type="qwen3vl_8b")
class Ideogram4Qwen3VLTEModel(sd1_clip.SD1ClipModel):
def __init__(self, device="cpu", dtype=None, model_options={}):
super().__init__(device=device, dtype=dtype, name="qwen3vl_8b", clip_model=Ideogram4Qwen3VLClipModel, model_options=model_options)
def encode_token_weights(self, token_weight_pairs):
out, pooled, extra = super().encode_token_weights(token_weight_pairs)
b, n, seq, h = out.shape # (B, n_taps=13, seq, 4096), ascending layer order.
out = out.permute(0, 2, 3, 1).reshape(b, seq, h * n) # (B, seq, 4096*13 = 53248).
return out, pooled, extra
class Ideogram4Qwen3VLTokenizer(comfy.text_encoders.qwen3vl.Qwen3VLTokenizer):
def __init__(self, embedding_directory=None, tokenizer_data={}):
super().__init__(embedding_directory=embedding_directory, tokenizer_data=tokenizer_data, model_type="qwen3vl_8b")
def tokenize_with_weights(self, text, return_word_ids=False, llama_template=None, images=[], prevent_empty_text=False, thinking=True, **kwargs):
# Ideogram 4 conditions on the no-think template; default thinking=True drops the empty think block qwen3vl adds.
return super().tokenize_with_weights(text, return_word_ids=return_word_ids, llama_template=llama_template, images=images, prevent_empty_text=prevent_empty_text, thinking=thinking, **kwargs)
def te_qwen3vl(dtype_llama=None, llama_quantization_metadata=None):
class Ideogram4Qwen3VLTEModel_(Ideogram4Qwen3VLTEModel):
def __init__(self, device="cpu", dtype=None, model_options={}):
if dtype_llama is not None:
dtype = dtype_llama
if llama_quantization_metadata is not None:
model_options = model_options.copy()
model_options["quantization_metadata"] = llama_quantization_metadata
super().__init__(device=device, dtype=dtype, model_options=model_options)
return Ideogram4Qwen3VLTEModel_

35
comfy/text_encoders/llama.py
View File

@ -251,6 +251,19 @@ class Qwen3_8BConfig:
lm_head: bool = True
stop_tokens = [151643, 151645]
@dataclass
class Qwen3VL_8BConfig(Qwen3_8BConfig):
max_position_embeddings: int = 262144
rope_theta: float = 5000000.0
rope_dims = [24, 20, 20]
interleaved_mrope = True
@dataclass
class Qwen3VL_4BConfig(Qwen3VL_8BConfig):
hidden_size: int = 2560
intermediate_size: int = 9728
lm_head: bool = False # 4B ties word embeddings
@dataclass
class Ovis25_2BConfig:
vocab_size: int = 151936
@ -703,7 +716,8 @@ class Llama2_(nn.Module):
interleaved_mrope=getattr(self.config, "interleaved_mrope", False),
device=device)
def forward(self, x, attention_mask=None, embeds=None, num_tokens=None, intermediate_output=None, final_layer_norm_intermediate=True, dtype=None, position_ids=None, embeds_info=[], past_key_values=None, input_ids=None):
def forward(self, x, attention_mask=None, embeds=None, num_tokens=None, intermediate_output=None, final_layer_norm_intermediate=True,
dtype=None, position_ids=None, embeds_info=[], past_key_values=None, input_ids=None,deepstack_embeds=None, visual_pos_masks=None):
if embeds is not None:
x = embeds
else:
@ -767,6 +781,10 @@ class Llama2_(nn.Module):
if current_kv is not None:
next_key_values.append(current_kv)
# DeepStack: add per-layer visual features into the first len() decoder layers at image positions (Qwen3-VL)
if deepstack_embeds is not None and i < len(deepstack_embeds):
x[visual_pos_masks] = x[visual_pos_masks] + deepstack_embeds[i].to(x)
if i == intermediate_output:
intermediate = x.clone()
@ -860,7 +878,7 @@ class BaseGenerate:
torch.empty([batch, model_config.num_key_value_heads, max_cache_len, model_config.head_dim], device=device, dtype=execution_dtype), 0))
return past_key_values
def generate(self, embeds=None, do_sample=True, max_length=256, temperature=1.0, top_k=50, top_p=0.9, min_p=0.0, repetition_penalty=1.0, seed=42, stop_tokens=None, initial_tokens=[], execution_dtype=None, min_tokens=0, presence_penalty=0.0, initial_input_ids=None):
def generate(self, embeds=None, do_sample=True, max_length=256, temperature=1.0, top_k=50, top_p=0.9, min_p=0.0, repetition_penalty=1.0, seed=42, stop_tokens=None, initial_tokens=[], execution_dtype=None, min_tokens=0, presence_penalty=0.0, initial_input_ids=None, position_ids=None, deepstack_embeds=None, visual_pos_masks=None):
device = embeds.device
if stop_tokens is None:
@ -884,10 +902,18 @@ class BaseGenerate:
generated_token_ids = []
pbar = comfy.utils.ProgressBar(max_length)
# MRoPE: prefill uses explicit 3D position_ids, decode continues from the last position
next_pos = int(position_ids[:, -1].max()) + 1 if position_ids is not None else None
# Generation loop
current_input_ids = initial_input_ids
for step in tqdm(range(max_length), desc="Generating tokens"):
x, _, past_key_values = self.model.forward(None, embeds=embeds, attention_mask=None, past_key_values=past_key_values, input_ids=current_input_ids)
# DeepStack visual features are injected on the prefill only; gemma4's forward lacks these kwargs.
extra = {}
if step == 0 and deepstack_embeds is not None:
extra["deepstack_embeds"] = deepstack_embeds
extra["visual_pos_masks"] = visual_pos_masks
x, _, past_key_values = self.model.forward(None, embeds=embeds, attention_mask=None, past_key_values=past_key_values, input_ids=current_input_ids, position_ids=position_ids, **extra)
logits = self.logits(x)[:, -1]
next_token = self.sample_token(logits, temperature, top_k, top_p, min_p, repetition_penalty, initial_tokens + generated_token_ids, generator, do_sample=do_sample, presence_penalty=presence_penalty)
token_id = next_token[0].item()
@ -895,6 +921,9 @@ class BaseGenerate:
embeds = self.model.embed_tokens(next_token).to(execution_dtype)
current_input_ids = next_token if initial_input_ids is not None else None
if next_pos is not None: # advance MRoPE position for the next (decode) step
position_ids = torch.tensor([[next_pos]], device=device)
next_pos += 1
pbar.update(1)
if token_id in stop_tokens:

35
comfy/text_encoders/qwen35.py
View File

@ -3,7 +3,6 @@ import torch.nn as nn
import torch.nn.functional as F
from dataclasses import dataclass, field
import os
import math
import comfy.model_management
from comfy.ldm.modules.attention import optimized_attention_for_device
@ -563,6 +562,8 @@ class Qwen35VisionModel(nn.Module):
for _ in range(config["depth"])
])
self.merger = Qwen35VisionPatchMerger(self.hidden_size, self.spatial_merge_size, config["out_hidden_size"], device=device, dtype=dtype, ops=ops)
self.deepstack_visual_indexes = [] # DeepStack, per-layer visual features (Qwen3-VL)
self.deepstack_merger_list = None
def rot_pos_emb(self, grid_thw):
merge_size = self.spatial_merge_size
@ -664,9 +665,14 @@ class Qwen35VisionModel(nn.Module):
).cumsum(dim=0, dtype=torch.int32)
cu_seqlens = F.pad(cu_seqlens, (1, 0), value=0)
optimized_attention = optimized_attention_for_device(x.device, mask=False, small_input=True)
for blk in self.blocks:
deepstack_features = []
for layer_num, blk in enumerate(self.blocks):
x = blk(x, cu_seqlens=cu_seqlens, position_embeddings=position_embeddings, optimized_attention=optimized_attention)
if self.deepstack_merger_list is not None and layer_num in self.deepstack_visual_indexes:
deepstack_features.append(self.deepstack_merger_list[self.deepstack_visual_indexes.index(layer_num)](x))
merged = self.merger(x)
if self.deepstack_merger_list is not None:
return merged, deepstack_features
return merged
# Model Wrapper
@ -690,30 +696,7 @@ class Qwen35(BaseLlama, BaseGenerate, torch.nn.Module):
return None, None
def forward(self, x, attention_mask=None, embeds=None, num_tokens=None, intermediate_output=None, final_layer_norm_intermediate=True, dtype=None, embeds_info=[], past_key_values=None):
grid = None
position_ids = None
offset = 0
for e in embeds_info:
if e.get("type") == "image":
grid = e.get("extra", None)
start = e.get("index")
if position_ids is None:
position_ids = torch.zeros((3, embeds.shape[1]), device=embeds.device)
position_ids[:, :start] = torch.arange(0, start, device=embeds.device)
end = e.get("size") + start
len_max = int(grid.max()) // 2
start_next = len_max + start
position_ids[:, end:] = torch.arange(start_next + offset, start_next + (embeds.shape[1] - end) + offset, device=embeds.device)
position_ids[0, start:end] = start + offset
max_d = int(grid[0][1]) // 2
position_ids[1, start:end] = torch.arange(start + offset, start + max_d + offset, device=embeds.device).unsqueeze(1).repeat(1, math.ceil((end - start) / max_d)).flatten(0)[:end - start]
max_d = int(grid[0][2]) // 2
position_ids[2, start:end] = torch.arange(start + offset, start + max_d + offset, device=embeds.device).unsqueeze(0).repeat(math.ceil((end - start) / max_d), 1).flatten(0)[:end - start]
offset += len_max - (end - start)
if grid is None:
position_ids = None
position_ids = comfy.text_encoders.qwen_vl.qwen2vl_mrope_position_ids(embeds_info, embeds.shape[1], embeds.device)
return super().forward(x, attention_mask=attention_mask, embeds=embeds, num_tokens=num_tokens, intermediate_output=intermediate_output, final_layer_norm_intermediate=final_layer_norm_intermediate, dtype=dtype, position_ids=position_ids, past_key_values=past_key_values)
def init_kv_cache(self, batch, max_cache_len, device, execution_dtype):

193
comfy/text_encoders/qwen3vl.py Normal file
View File

@ -0,0 +1,193 @@
import os
import torch
import torch.nn as nn
import torch.nn.functional as F
from transformers import Qwen2Tokenizer
from comfy import sd1_clip
import comfy.text_encoders.qwen_vl
from .qwen35 import Qwen35VisionModel
from .llama import BaseLlama, BaseQwen3, BaseGenerate, Llama2_, Qwen3VL_4BConfig, Qwen3VL_8BConfig
QWEN3VL_VISION = {
"qwen3vl_4b": dict(hidden_size=1024, intermediate_size=4096, depth=24, deepstack_visual_indexes=[5, 11, 17]),
"qwen3vl_8b": dict(hidden_size=1152, intermediate_size=4304, depth=27, deepstack_visual_indexes=[8, 16, 24]),
}
QWEN3VL_VISION_COMMON = dict(num_heads=16, patch_size=16, temporal_patch_size=2, in_channels=3,
spatial_merge_size=2, num_position_embeddings=2304)
QWEN3VL_CONFIGS = {"qwen3vl_4b": Qwen3VL_4BConfig, "qwen3vl_8b": Qwen3VL_8BConfig}
class Qwen3VLDeepstackMerger(nn.Module):
# DeepStack merger: postshuffle LayerNorm (applied after spatial merge), unlike the main merger.
def __init__(self, hidden_size, spatial_merge_size, out_hidden_size, device=None, dtype=None, ops=None):
super().__init__()
self.merge_dim = hidden_size * (spatial_merge_size ** 2)
self.norm = ops.LayerNorm(self.merge_dim, eps=1e-6, device=device, dtype=dtype)
self.linear_fc1 = ops.Linear(self.merge_dim, self.merge_dim, device=device, dtype=dtype)
self.linear_fc2 = ops.Linear(self.merge_dim, out_hidden_size, device=device, dtype=dtype)
def forward(self, x):
x = self.norm(x.view(-1, self.merge_dim))
return self.linear_fc2(F.gelu(self.linear_fc1(x)))
class Qwen3VLVisionModel(Qwen35VisionModel):
# Qwen3.5 vision + DeepStack
def __init__(self, config, device=None, dtype=None, ops=None):
super().__init__(config, device=device, dtype=dtype, ops=ops)
self.deepstack_visual_indexes = config["deepstack_visual_indexes"]
self.deepstack_merger_list = nn.ModuleList([
Qwen3VLDeepstackMerger(self.hidden_size, self.spatial_merge_size, config["out_hidden_size"], device=device, dtype=dtype, ops=ops)
for _ in self.deepstack_visual_indexes
])
class Qwen3VL(BaseLlama, BaseQwen3, BaseGenerate, torch.nn.Module):
model_type = "qwen3vl_8b"
def __init__(self, config_dict, dtype, device, operations):
super().__init__()
config = QWEN3VL_CONFIGS[self.model_type](**config_dict)
self.num_layers = config.num_hidden_layers
self.model = Llama2_(config, device=device, dtype=dtype, ops=operations)
vision_config = {**QWEN3VL_VISION_COMMON, **QWEN3VL_VISION[self.model_type], "out_hidden_size": config.hidden_size}
self.visual = Qwen3VLVisionModel(vision_config, device=device, dtype=dtype, ops=operations)
self.dtype = dtype
def preprocess_embed(self, embed, device):
if embed["type"] == "image":
# Qwen3-VL normalizes to [-1, 1] (mean/std 0.5), unlike Qwen2.5-VL's CLIP normalization.
image, grid = comfy.text_encoders.qwen_vl.process_qwen2vl_images(embed["data"], patch_size=16, image_mean=[0.5, 0.5, 0.5], image_std=[0.5, 0.5, 0.5])
merged, deepstack = self.visual(image.to(device, dtype=torch.float32), grid)
return merged, {"grid": grid, "deepstack": deepstack}
return None, None
def build_image_inputs(self, embeds, embeds_info):
# Returns (position_ids, visual_pos_masks, deepstack) for the prompt
images = sorted([e for e in embeds_info if e.get("type") == "image"], key=lambda e: e["index"])
if len(images) == 0:
return None, None, None
device = embeds.device
seq = embeds.shape[1]
position_ids = comfy.text_encoders.qwen_vl.qwen2vl_mrope_position_ids(embeds_info, seq, device)
# DeepStack: mask of image positions + per-vision-layer features to inject there.
visual_pos_masks = torch.zeros((1, seq), dtype=torch.bool, device=device)
deepstack = None
for e in images:
start = e["index"]
end = e["size"] + start
visual_pos_masks[0, start:end] = True
ds = e["extra"]["deepstack"]
if deepstack is None:
deepstack = [d for d in ds]
else:
deepstack = [torch.cat([deepstack[i], ds[i]], dim=0) for i in range(len(ds))]
return position_ids, visual_pos_masks, deepstack
def _make_qwen3vl_model(model_type):
class Qwen3VL_(Qwen3VL):
pass
Qwen3VL_.model_type = model_type
return Qwen3VL_
class Qwen3VLClipModel(sd1_clip.SDClipModel):
def __init__(self, device="cpu", layer="hidden", layer_idx=-1, dtype=None, attention_mask=True, model_options={}, model_type="qwen3vl_8b"):
super().__init__(device=device, layer=layer, layer_idx=layer_idx, textmodel_json_config={},
dtype=dtype, special_tokens={"pad": 151643}, layer_norm_hidden_state=False,
model_class=_make_qwen3vl_model(model_type), enable_attention_masks=attention_mask,
return_attention_masks=attention_mask, model_options=model_options)
def generate(self, tokens, do_sample, max_length, temperature, top_k, top_p, min_p, repetition_penalty, seed, presence_penalty=0.0):
if isinstance(tokens, dict):
tokens = next(iter(tokens.values()))
tokens_only = [[t[0] for t in b] for b in tokens]
embeds, _, _, embeds_info = self.process_tokens(tokens_only, self.execution_device)
position_ids, visual_pos_masks, deepstack = self.transformer.build_image_inputs(embeds, embeds_info)
return self.transformer.generate(embeds, do_sample, max_length, temperature, top_k, top_p, min_p, repetition_penalty, seed,
presence_penalty=presence_penalty, position_ids=position_ids,
visual_pos_masks=visual_pos_masks, deepstack_embeds=deepstack)
class Qwen3VLTEModel(sd1_clip.SD1ClipModel):
def __init__(self, device="cpu", dtype=None, model_options={}, model_type="qwen3vl_8b"):
clip_model = lambda **kw: Qwen3VLClipModel(**kw, model_type=model_type)
super().__init__(device=device, dtype=dtype, name=model_type, clip_model=clip_model, model_options=model_options)
class Qwen3VLSDTokenizer(sd1_clip.SDTokenizer):
def __init__(self, embedding_directory=None, tokenizer_data={}, embedding_size=4096, embedding_key="qwen3vl_8b"):
tokenizer_path = os.path.join(os.path.dirname(os.path.realpath(__file__)), "qwen25_tokenizer")
super().__init__(tokenizer_path, pad_with_end=False, embedding_directory=embedding_directory, embedding_size=embedding_size, embedding_key=embedding_key, tokenizer_class=Qwen2Tokenizer,
has_start_token=False, has_end_token=False, pad_to_max_length=False, max_length=99999999, min_length=1, pad_token=151643, tokenizer_data=tokenizer_data)
class Qwen3VLTokenizer(sd1_clip.SD1Tokenizer):
def __init__(self, embedding_directory=None, tokenizer_data={}, model_type="qwen3vl_8b"):
embedding_size = 2560 if model_type == "qwen3vl_4b" else 4096
tokenizer = lambda *a, **kw: Qwen3VLSDTokenizer(*a, **kw, embedding_size=embedding_size, embedding_key=model_type)
super().__init__(embedding_directory=embedding_directory, tokenizer_data=tokenizer_data, name=model_type, tokenizer=tokenizer)
self.llama_template = "<|im_start|>user\n{}<|im_end|>\n<|im_start|>assistant\n"
self.llama_template_images = "<|im_start|>user\n<|vision_start|><|image_pad|><|vision_end|>{}<|im_end|>\n<|im_start|>assistant\n"
def tokenize_with_weights(self, text, return_word_ids=False, llama_template=None, images=[], prevent_empty_text=False, thinking=False, **kwargs):
image = kwargs.get("image", None)
if image is not None and len(images) == 0:
images = [image[i:i + 1] for i in range(image.shape[0])]
skip_template = text.startswith('<|im_start|>')
if prevent_empty_text and text == '':
text = ' '
if skip_template:
llama_text = text
else:
if llama_template is not None:
template = llama_template
elif len(images) == 0:
template = self.llama_template
else:
template = self.llama_template_images
if len(images) > 1:
vision_block = "<|vision_start|><|image_pad|><|vision_end|>"
template = template.replace(vision_block, vision_block * len(images), 1)
llama_text = template.format(text)
if not thinking: # Qwen3 convention: empty think block suppresses reasoning
llama_text += "<think>\n\n</think>\n\n"
tokens = super().tokenize_with_weights(llama_text, return_word_ids=return_word_ids, disable_weights=True, **kwargs)
key_name = next(iter(tokens))
embed_count = 0
for r in tokens[key_name]:
for i in range(len(r)):
if r[i][0] == 151655: # <|image_pad|>
if len(images) > embed_count:
r[i] = ({"type": "image", "data": images[embed_count], "original_type": "image"},) + r[i][1:]
embed_count += 1
return tokens
def tokenizer(model_type="qwen3vl_8b"):
class Qwen3VLTokenizer_(Qwen3VLTokenizer):
def __init__(self, embedding_directory=None, tokenizer_data={}):
super().__init__(embedding_directory=embedding_directory, tokenizer_data=tokenizer_data, model_type=model_type)
return Qwen3VLTokenizer_
def te(dtype_llama=None, llama_quantization_metadata=None, model_type="qwen3vl_8b"):
class Qwen3VLTEModel_(Qwen3VLTEModel):
def __init__(self, device="cpu", dtype=None, model_options={}):
if dtype_llama is not None:
dtype = dtype_llama
if llama_quantization_metadata is not None:
model_options = model_options.copy()
model_options["quantization_metadata"] = llama_quantization_metadata
super().__init__(device=device, dtype=dtype, model_options=model_options, model_type=model_type)
return Qwen3VLTEModel_

26
comfy/text_encoders/qwen_vl.py
View File

@ -88,6 +88,32 @@ def process_qwen2vl_images(
return flatten_patches, image_grid_thw
def qwen2vl_mrope_position_ids(embeds_info, seq_len, device):
# (3, seq_len) T/H/W MRoPE position ids: text runs sequentially, each image span gets its grid positions.
# Returns None when there are no image embeds. `extra` is the image grid_thw, or a dict carrying it under "grid".
position_ids = None
offset = 0
for e in embeds_info:
if e.get("type") == "image":
extra = e.get("extra", None)
grid = extra["grid"] if isinstance(extra, dict) else extra
start = e.get("index")
if position_ids is None:
position_ids = torch.zeros((3, seq_len), device=device)
position_ids[:, :start] = torch.arange(0, start, device=device)
end = e.get("size") + start
len_max = int(grid.max()) // 2
start_next = len_max + start
position_ids[:, end:] = torch.arange(start_next + offset, start_next + (seq_len - end) + offset, device=device)
position_ids[0, start:end] = start + offset
max_d = int(grid[0][1]) // 2
position_ids[1, start:end] = torch.arange(start + offset, start + max_d + offset, device=device).unsqueeze(1).repeat(1, math.ceil((end - start) / max_d)).flatten(0)[:end - start]
max_d = int(grid[0][2]) // 2
position_ids[2, start:end] = torch.arange(start + offset, start + max_d + offset, device=device).unsqueeze(0).repeat(math.ceil((end - start) / max_d), 1).flatten(0)[:end - start]
offset += len_max - (end - start)
return position_ids
class VisionPatchEmbed(nn.Module):
def __init__(
self,

12
comfy_extras/nodes_ace.py
View File

@ -11,7 +11,7 @@ class TextEncodeAceStepAudio(IO.ComfyNode):
def define_schema(cls):
return IO.Schema(
node_id="TextEncodeAceStepAudio",
category="model/conditioning",
category="model/conditioning/ace",
inputs=[
IO.Clip.Input("clip"),
IO.String.Input("tags", multiline=True, dynamic_prompts=True),
@ -33,7 +33,7 @@ class TextEncodeAceStepAudio15(IO.ComfyNode):
def define_schema(cls):
return IO.Schema(
node_id="TextEncodeAceStepAudio1.5",
category="model/conditioning",
category="model/conditioning/ace",
inputs=[
IO.Clip.Input("clip"),
IO.String.Input("tags", multiline=True, dynamic_prompts=True),
@ -67,7 +67,7 @@ class EmptyAceStepLatentAudio(IO.ComfyNode):
return IO.Schema(
node_id="EmptyAceStepLatentAudio",
display_name="Empty Ace Step 1.0 Latent Audio",
category="model/latent/audio",
category="model/latent/ace",
inputs=[
IO.Float.Input("seconds", default=120.0, min=1.0, max=1000.0, step=0.1),
IO.Int.Input(
@ -90,7 +90,7 @@ class EmptyAceStep15LatentAudio(IO.ComfyNode):
return IO.Schema(
node_id="EmptyAceStep1.5LatentAudio",
display_name="Empty Ace Step 1.5 Latent Audio",
category="model/latent/audio",
category="model/latent/ace",
inputs=[
IO.Float.Input("seconds", default=120.0, min=1.0, max=1000.0, step=0.01),
IO.Int.Input(
@ -111,8 +111,8 @@ class ReferenceAudio(IO.ComfyNode):
def define_schema(cls):
return IO.Schema(
node_id="ReferenceTimbreAudio",
display_name="Reference Audio",
category="advanced/conditioning/audio",
display_name="Set Reference Audio",
category="model/conditioning",
is_experimental=True,
description="This node sets the reference audio for ace step 1.5",
inputs=[

2
comfy_extras/nodes_apg.py
View File

@ -16,7 +16,7 @@ class APG(io.ComfyNode):
return io.Schema(
node_id="APG",
display_name="Adaptive Projected Guidance",
category="model/sampling/custom_sampling",
category="model/sampling/custom",
inputs=[
io.Model.Input("model"),
io.Float.Input(

4
comfy_extras/nodes_ar_video.py
View File

@ -19,7 +19,7 @@ class EmptyARVideoLatent(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="EmptyARVideoLatent",
category="model/latent/video",
category="model/latent/autoregressive",
inputs=[
io.Int.Input("width", default=832, min=16, max=8192, step=16),
io.Int.Input("height", default=480, min=16, max=8192, step=16),
@ -85,7 +85,7 @@ class ARVideoI2V(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="ARVideoI2V",
category="model/conditioning/video_models",
category="model/conditioning/autoregressive",
inputs=[
io.Model.Input("model"),
io.Vae.Input("vae"),

10
comfy_extras/nodes_audio.py
View File

@ -16,7 +16,7 @@ class EmptyLatentAudio(IO.ComfyNode):
return IO.Schema(
node_id="EmptyLatentAudio",
display_name="Empty Latent Audio",
category="model/latent/audio",
category="model/latent",
essentials_category="Audio",
inputs=[
IO.Float.Input("seconds", default=47.6, min=1.0, max=1000.0, step=0.1),
@ -41,7 +41,7 @@ class ConditioningStableAudio(IO.ComfyNode):
def define_schema(cls):
return IO.Schema(
node_id="ConditioningStableAudio",
category="model/conditioning",
category="model/conditioning/stable audio",
inputs=[
IO.Conditioning.Input("positive"),
IO.Conditioning.Input("negative"),
@ -70,7 +70,7 @@ class VAEEncodeAudio(IO.ComfyNode):
node_id="VAEEncodeAudio",
search_aliases=["audio to latent"],
display_name="VAE Encode Audio",
category="model/latent/audio",
category="model/latent",
inputs=[
IO.Audio.Input("audio"),
IO.Vae.Input("vae"),
@ -115,7 +115,7 @@ class VAEDecodeAudio(IO.ComfyNode):
node_id="VAEDecodeAudio",
search_aliases=["latent to audio"],
display_name="VAE Decode Audio",
category="model/latent/audio",
category="model/latent",
inputs=[
IO.Latent.Input("samples"),
IO.Vae.Input("vae"),
@ -137,7 +137,7 @@ class VAEDecodeAudioTiled(IO.ComfyNode):
node_id="VAEDecodeAudioTiled",
search_aliases=["latent to audio"],
display_name="VAE Decode Audio (Tiled)",
category="model/latent/audio",
category="model/latent",
inputs=[
IO.Latent.Input("samples"),
IO.Vae.Input("vae"),

23
comfy_extras/nodes_bernini.py
View File

@ -39,9 +39,9 @@ class BerniniConditioning(io.ComfyNode):
return io.Schema(
node_id="BerniniConditioning",
display_name="Bernini Conditioning",
category="conditioning/video_models",
category="model/conditioning/bernini",
description="Conditioning node for Bernini in-context video/image conditioning. It can be used for the following tasks: t2v (text-to-video), v2v (video-to-video), rv2v (reference-guided video editing), r2v (reference-to-video), ads2v (insert image/video into video)."
"Reference images injected as in-context tokens (r2v, rv2v) are encoded independently at their own native aspect ratio (long edge capped at ref_max_size)",
"Reference images injected as in-context tokens (r2v, rv2v) are encoded independently at their own native aspect ratio (long edge capped at ref_max_size)",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -50,14 +50,11 @@ class BerniniConditioning(io.ComfyNode):
io.Int.Input("height", default=480, min=16, max=8192, step=16),
io.Int.Input("length", default=81, min=1, max=8192, step=4),
io.Int.Input("batch_size", default=1, min=1, max=4096),
io.Image.Input("source_video", optional=True, tooltip=(
"Source video to edit or restyle (v2v, rv2v). Resized to width/height and trimmed to length.")),
io.Image.Input("reference_video", optional=True, tooltip=(
"Video to insert into the source video (ads2v).")),
io.Image.Input("source_video", optional=True, tooltip=("Source video to edit or restyle (v2v, rv2v). Resized to width/height and trimmed to length.")),
io.Image.Input("reference_video", optional=True, tooltip=("Video to insert into the source video (ads2v).")),
io.Autogrow.Input("reference_images", optional=True,
template=io.Autogrow.TemplatePrefix(
input=io.Image.Input("reference_image", tooltip=(
"Reference image injected as an in-context token (r2v, rv2v).")),
input=io.Image.Input("reference_image", tooltip=("Reference image injected as an in-context token (r2v, rv2v).")),
prefix="reference_image_", min=0, max=8)),
io.Int.Input("ref_max_size", default=848, min=16, max=8192, step=16, optional=True, tooltip=(
"Max size for the long edge of reference_video and reference_images. Resized with preserved aspect ratio and snapped to 16px.")),
@ -70,10 +67,8 @@ class BerniniConditioning(io.ComfyNode):
)
@classmethod
def execute(cls, positive, negative, vae, width, height, length, batch_size,
source_video=None, reference_video=None, reference_images=None, ref_max_size=848) -> io.NodeOutput:
latent = torch.zeros([batch_size, 16, ((length - 1) // 4) + 1, height // 8, width // 8],
device=comfy.model_management.intermediate_device())
def execute(cls, positive, negative, vae, width, height, length, batch_size, source_video=None, reference_video=None, reference_images=None, ref_max_size=848) -> io.NodeOutput:
latent = torch.zeros([batch_size, 16, ((length - 1) // 4) + 1, height // 8, width // 8], device=comfy.model_management.intermediate_device())
# source_video (1), reference_video (2), reference_images (3, 4, ...).
context = []
@ -106,9 +101,7 @@ class BerniniConditioning(io.ComfyNode):
class BerniniExtension(ComfyExtension):
@override
async def get_node_list(self) -> list[type[io.ComfyNode]]:
return [
BerniniConditioning,
]
return [BerniniConditioning,]
async def comfy_entrypoint() -> BerniniExtension:

2
comfy_extras/nodes_camera_trajectory.py
View File

@ -153,7 +153,7 @@ class WanCameraEmbedding(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="WanCameraEmbedding",
category="model/conditioning/video_models",
category="model/conditioning/wan/camera",
inputs=[
io.Combo.Input(
"camera_pose",

4
comfy_extras/nodes_chroma_radiance.py
View File

@ -13,7 +13,7 @@ class EmptyChromaRadianceLatentImage(io.ComfyNode):
def define_schema(cls) -> io.Schema:
return io.Schema(
node_id="EmptyChromaRadianceLatentImage",
category="model/latent/chroma_radiance",
category="model/latent/chroma radiance",
inputs=[
io.Int.Input(id="width", default=1024, min=16, max=nodes.MAX_RESOLUTION, step=16),
io.Int.Input(id="height", default=1024, min=16, max=nodes.MAX_RESOLUTION, step=16),
@ -33,7 +33,7 @@ class ChromaRadianceOptions(io.ComfyNode):
def define_schema(cls) -> io.Schema:
return io.Schema(
node_id="ChromaRadianceOptions",
category="model/patch/chroma_radiance",
category="model/patch/chroma radiance",
description="Allows setting advanced options for the Chroma Radiance model.",
inputs=[
io.Model.Input(id="model"),

6
comfy_extras/nodes_clip_sdxl.py
View File

@ -9,7 +9,8 @@ class CLIPTextEncodeSDXLRefiner(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="CLIPTextEncodeSDXLRefiner",
category="advanced/conditioning",
display_name="CLIP Text Encode (SDXL Refiner)",
category="model/conditioning/stable diffusion",
inputs=[
io.Float.Input("ascore", default=6.0, min=0.0, max=1000.0, step=0.01),
io.Int.Input("width", default=1024, min=0, max=nodes.MAX_RESOLUTION),
@ -30,7 +31,8 @@ class CLIPTextEncodeSDXL(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="CLIPTextEncodeSDXL",
category="advanced/conditioning",
display_name="CLIP Text Encode (SDXL)",
category="model/conditioning/stable diffusion",
inputs=[
io.Clip.Input("clip"),
io.Int.Input("width", default=1024, min=0, max=nodes.MAX_RESOLUTION),

1
comfy_extras/nodes_context_windows.py
View File

@ -66,6 +66,7 @@ class WanContextWindowsManualNode(ContextWindowsManualNode):
schema.node_id = "WanContextWindowsManual"
schema.display_name = "WAN Context Windows (Manual)"
schema.description = "Manually set context windows for WAN-like models (dim=2)."
schema.category="model/patch/wan"
schema.inputs = [
io.Model.Input("model", tooltip="The model to apply context windows to during sampling."),
io.Int.Input("context_length", min=1, max=nodes.MAX_RESOLUTION, step=4, default=81, tooltip="The length of the context window.", advanced=True),

3
comfy_extras/nodes_controlnet.py
View File

@ -9,6 +9,8 @@ class SetUnionControlNetType(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="SetUnionControlNetType",
search_aliases=["set controlnet type", "union controlnet type"],
display_name="Set Union ControlNet Type",
category="model/conditioning/controlnet",
inputs=[
io.ControlNet.Input("control_net"),
@ -39,6 +41,7 @@ class ControlNetInpaintingAliMamaApply(io.ComfyNode):
return io.Schema(
node_id="ControlNetInpaintingAliMamaApply",
search_aliases=["masked controlnet"],
display_name="Apply ControlNet Inpainting (AliMama)",
category="model/conditioning/controlnet",
inputs=[
io.Conditioning.Input("positive"),

6
comfy_extras/nodes_cosmos.py
View File

@ -13,7 +13,7 @@ class EmptyCosmosLatentVideo(io.ComfyNode):
def define_schema(cls) -> io.Schema:
return io.Schema(
node_id="EmptyCosmosLatentVideo",
category="model/latent/video",
category="model/latent/cosmos",
inputs=[
io.Int.Input("width", default=1280, min=16, max=nodes.MAX_RESOLUTION, step=16),
io.Int.Input("height", default=704, min=16, max=nodes.MAX_RESOLUTION, step=16),
@ -45,7 +45,7 @@ class CosmosImageToVideoLatent(io.ComfyNode):
def define_schema(cls) -> io.Schema:
return io.Schema(
node_id="CosmosImageToVideoLatent",
category="model/conditioning/inpaint",
category="model/conditioning/cosmos",
inputs=[
io.Vae.Input("vae"),
io.Int.Input("width", default=1280, min=16, max=nodes.MAX_RESOLUTION, step=16),
@ -88,7 +88,7 @@ class CosmosPredict2ImageToVideoLatent(io.ComfyNode):
def define_schema(cls) -> io.Schema:
return io.Schema(
node_id="CosmosPredict2ImageToVideoLatent",
category="model/conditioning/inpaint",
category="model/conditioning/cosmos",
inputs=[
io.Vae.Input("vae"),
io.Int.Input("width", default=848, min=16, max=nodes.MAX_RESOLUTION, step=16),

6
comfy_extras/nodes_custom_sampler.py
View File

@ -729,7 +729,7 @@ class SamplerCustom(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="SamplerCustom",
category="model/sampling/custom_sampling",
category="model/sampling/custom",
inputs=[
io.Model.Input("model"),
io.Boolean.Input("add_noise", default=True, advanced=True),
@ -1015,7 +1015,7 @@ class SamplerCustomAdvanced(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="SamplerCustomAdvanced",
category="model/sampling/custom_sampling",
category="model/sampling/custom",
inputs=[
io.Noise.Input("noise"),
io.Guider.Input("guider"),
@ -1143,7 +1143,7 @@ class CFGOverride(io.ComfyNode):
display_name="CFG Override",
description="Override cfg to a fixed value over a [start, end] percent (sigma) range. "
"With multiple overrides, the one nearest the sampler wins on overlap.",
category="sampling/custom_sampling",
category="model/sampling/guiders",
inputs=[
io.Model.Input("model"),
io.Float.Input("cfg", default=1.0, min=0.0, max=100.0, step=0.1, round=0.01),

4
comfy_extras/nodes_easycache.py
View File

@ -363,7 +363,7 @@ class EasyCacheNode(io.ComfyNode):
node_id="EasyCache",
display_name="EasyCache",
description="Native EasyCache implementation.",
category="advanced/debug/model",
category="advanced/debug",
is_experimental=True,
inputs=[
io.Model.Input("model", tooltip="The model to add EasyCache to."),
@ -496,7 +496,7 @@ class LazyCacheNode(io.ComfyNode):
node_id="LazyCache",
display_name="LazyCache",
description="A homebrew version of EasyCache - even 'easier' version of EasyCache to implement. Overall works worse than EasyCache, but better in some rare cases AND universal compatibility with everything in ComfyUI.",
category="advanced/debug/model",
category="advanced/debug",
is_experimental=True,
inputs=[
io.Model.Input("model", tooltip="The model to add LazyCache to."),

3
comfy_extras/nodes_edit_model.py
View File

@ -8,7 +8,8 @@ class ReferenceLatent(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="ReferenceLatent",
category="advanced/conditioning/edit_models",
display_name="Set Reference Latent",
category="model/conditioning",
description="This node sets the guiding latent for an edit model. If the model supports it you can chain multiple to set multiple reference images.",
inputs=[
io.Conditioning.Input("conditioning"),

12
comfy_extras/nodes_flux.py
View File

@ -13,7 +13,7 @@ class CLIPTextEncodeFlux(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="CLIPTextEncodeFlux",
category="advanced/conditioning/flux",
category="model/conditioning/flux",
inputs=[
io.Clip.Input("clip"),
io.String.Input("clip_l", multiline=True, dynamic_prompts=True),
@ -40,7 +40,7 @@ class EmptyFlux2LatentImage(io.ComfyNode):
return io.Schema(
node_id="EmptyFlux2LatentImage",
display_name="Empty Flux 2 Latent",
category="model/latent",
category="model/latent/flux",
inputs=[
io.Int.Input("width", default=1024, min=16, max=nodes.MAX_RESOLUTION, step=16),
io.Int.Input("height", default=1024, min=16, max=nodes.MAX_RESOLUTION, step=16),
@ -61,7 +61,7 @@ class FluxGuidance(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="FluxGuidance",
category="advanced/conditioning/flux",
category="model/conditioning/flux",
inputs=[
io.Conditioning.Input("conditioning"),
io.Float.Input("guidance", default=3.5, min=0.0, max=100.0, step=0.1),
@ -84,7 +84,7 @@ class FluxDisableGuidance(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="FluxDisableGuidance",
category="advanced/conditioning/flux",
category="model/conditioning/flux",
description="This node completely disables the guidance embed on Flux and Flux like models",
inputs=[
io.Conditioning.Input("conditioning"),
@ -128,7 +128,7 @@ class FluxKontextImageScale(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="FluxKontextImageScale",
category="advanced/conditioning/flux",
category="model/conditioning/flux",
description="This node resizes the image to one that is more optimal for flux kontext.",
inputs=[
io.Image.Input("image"),
@ -156,7 +156,7 @@ class FluxKontextMultiReferenceLatentMethod(io.ComfyNode):
return io.Schema(
node_id="FluxKontextMultiReferenceLatentMethod",
display_name="Edit Model Reference Method",
category="advanced/conditioning/flux",
category="model/conditioning/flux",
inputs=[
io.Conditioning.Input("conditioning"),
io.Combo.Input(

8
comfy_extras/nodes_hidream.py
View File

@ -11,8 +11,9 @@ class QuadrupleCLIPLoader(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="QuadrupleCLIPLoader",
category="advanced/loaders",
description="[Recipes]\n\nhidream: long clip-l, long clip-g, t5xxl, llama_8b_3.1_instruct",
display_name="Load CLIP (Quadruple)",
category="model/loaders",
description="Recipes:\nhidream: long clip-l, long clip-g, t5xxl, llama_8b_3.1_instruct",
inputs=[
io.Combo.Input("clip_name1", options=folder_paths.get_filename_list("text_encoders")),
io.Combo.Input("clip_name2", options=folder_paths.get_filename_list("text_encoders")),
@ -38,8 +39,9 @@ class CLIPTextEncodeHiDream(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="CLIPTextEncodeHiDream",
display_name="CLIP Text Encode (HiDream)",
search_aliases=["hidream prompt"],
category="advanced/conditioning",
category="model/conditioning/hidream",
inputs=[
io.Clip.Input("clip"),
io.String.Input("clip_l", multiline=True, dynamic_prompts=True),

6
comfy_extras/nodes_hidream_o1.py
View File

@ -14,7 +14,7 @@ class EmptyHiDreamO1LatentImage(io.ComfyNode):
return io.Schema(
node_id="EmptyHiDreamO1LatentImage",
display_name="Empty HiDream-O1 Latent Image",
category="model/latent/image",
category="model/latent/hidream",
description=(
"Empty pixel-space latent for HiDream-O1-Image. The model was "
"trained at ~4 megapixels; lower resolutions go off-distribution "
@ -47,7 +47,7 @@ class HiDreamO1ReferenceImages(io.ComfyNode):
return io.Schema(
node_id="HiDreamO1ReferenceImages",
display_name="HiDream-O1 Reference Images",
category="model/conditioning/image",
category="model/conditioning/hidream",
description=(
"Attach 1-10 reference images to conditioning, one for edit instruction"
"or multiple for subject-driven personalization."
@ -117,7 +117,7 @@ class HiDreamO1PatchSeamSmoothing(io.ComfyNode):
return io.Schema(
node_id="HiDreamO1PatchSeamSmoothing",
display_name="HiDream-O1 Patch Seam Smoothing",
category="advanced/model",
category="model/patch/hidream",
is_experimental=True,
description=(
"Average the model output across multiple shifted patch-grid "

20
comfy_extras/nodes_hunyuan.py
View File

@ -14,7 +14,8 @@ class CLIPTextEncodeHunyuanDiT(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="CLIPTextEncodeHunyuanDiT",
category="advanced/conditioning",
display_name="CLIP Text Encode (Hunyuan Image)",
category="model/conditioning/hunyuan image",
inputs=[
io.Clip.Input("clip"),
io.String.Input("bert", multiline=True, dynamic_prompts=True),
@ -41,7 +42,7 @@ class EmptyHunyuanLatentVideo(io.ComfyNode):
return io.Schema(
node_id="EmptyHunyuanLatentVideo",
display_name="Empty HunyuanVideo 1.0 Latent",
category="model/latent/video",
category="model/latent/hunyuan video",
inputs=[
io.Int.Input("width", default=848, min=16, max=nodes.MAX_RESOLUTION, step=16),
io.Int.Input("height", default=480, min=16, max=nodes.MAX_RESOLUTION, step=16),
@ -67,6 +68,7 @@ class EmptyHunyuanVideo15Latent(EmptyHunyuanLatentVideo):
schema = super().define_schema()
schema.node_id = "EmptyHunyuanVideo15Latent"
schema.display_name = "Empty HunyuanVideo 1.5 Latent"
schema.category = "model/latent/hunyuan video"
return schema
@classmethod
@ -81,7 +83,7 @@ class HunyuanVideo15ImageToVideo(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="HunyuanVideo15ImageToVideo",
category="model/conditioning/video_models",
category="model/conditioning/hunyuan video",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -132,7 +134,7 @@ class HunyuanVideo15SuperResolution(io.ComfyNode):
return io.Schema(
node_id="HunyuanVideo15SuperResolution",
display_name="Hunyuan Video 1.5 Super Resolution",
category="model/conditioning/video_models",
category="model/conditioning/hunyuan video",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -227,7 +229,7 @@ class HunyuanVideo15LatentUpscaleWithModel(io.ComfyNode):
return io.Schema(
node_id="HunyuanVideo15LatentUpscaleWithModel",
display_name="Hunyuan Video 15 Latent Upscale With Model",
category="model/latent",
category="model/latent/hunyhuan video",
inputs=[
io.LatentUpscaleModel.Input("model"),
io.Latent.Input("samples"),
@ -276,7 +278,7 @@ class TextEncodeHunyuanVideo_ImageToVideo(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="TextEncodeHunyuanVideo_ImageToVideo",
category="advanced/conditioning",
category="model/conditioning/hunyuan video",
inputs=[
io.Clip.Input("clip"),
io.ClipVisionOutput.Input("clip_vision_output"),
@ -308,7 +310,7 @@ class HunyuanImageToVideo(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="HunyuanImageToVideo",
category="model/conditioning/video_models",
category="model/conditioning/hunyuan video",
inputs=[
io.Conditioning.Input("positive"),
io.Vae.Input("vae"),
@ -359,7 +361,7 @@ class EmptyHunyuanImageLatent(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="EmptyHunyuanImageLatent",
category="model/latent",
category="model/latent/hunyuan image",
inputs=[
io.Int.Input("width", default=2048, min=64, max=nodes.MAX_RESOLUTION, step=32),
io.Int.Input("height", default=2048, min=64, max=nodes.MAX_RESOLUTION, step=32),
@ -384,7 +386,7 @@ class HunyuanRefinerLatent(io.ComfyNode):
return io.Schema(
node_id="HunyuanRefinerLatent",
display_name="Hunyuan Latent Refiner",
category="model/conditioning/video_models",
category="model/conditioning/hunyuan video",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),

8
comfy_extras/nodes_hunyuan3d.py
View File

@ -12,7 +12,7 @@ class EmptyLatentHunyuan3Dv2(IO.ComfyNode):
def define_schema(cls):
return IO.Schema(
node_id="EmptyLatentHunyuan3Dv2",
category="model/latent/3d",
category="model/latent/hunyuan 3d",
inputs=[
IO.Int.Input("resolution", default=3072, min=1, max=8192),
IO.Int.Input("batch_size", default=1, min=1, max=4096, tooltip="The number of latent images in the batch."),
@ -35,7 +35,7 @@ class Hunyuan3Dv2Conditioning(IO.ComfyNode):
def define_schema(cls):
return IO.Schema(
node_id="Hunyuan3Dv2Conditioning",
category="model/conditioning/3d_models",
category="model/conditioning/hunyuan 3d",
inputs=[
IO.ClipVisionOutput.Input("clip_vision_output"),
],
@ -60,7 +60,7 @@ class Hunyuan3Dv2ConditioningMultiView(IO.ComfyNode):
def define_schema(cls):
return IO.Schema(
node_id="Hunyuan3Dv2ConditioningMultiView",
category="model/conditioning/3d_models",
category="model/conditioning/hunyuan 3d",
inputs=[
IO.ClipVisionOutput.Input("front", optional=True),
IO.ClipVisionOutput.Input("left", optional=True),
@ -97,7 +97,7 @@ class VAEDecodeHunyuan3D(IO.ComfyNode):
def define_schema(cls):
return IO.Schema(
node_id="VAEDecodeHunyuan3D",
category="model/latent/3d",
category="model/latent/hunyuan 3d",
inputs=[
IO.Latent.Input("samples"),
IO.Vae.Input("vae"),

2
comfy_extras/nodes_ideogram4.py
View File

@ -38,7 +38,7 @@ class Ideogram4Scheduler(io.ComfyNode):
return io.Schema(
node_id="Ideogram4Scheduler",
display_name="Ideogram 4 Scheduler",
category="sampling/custom_sampling/schedulers",
category="model/sampling/schedulers",
inputs=[
io.Int.Input("steps", default=20, min=1, max=200),
io.Int.Input("width", default=1024, min=256, max=8192, step=16),

7
comfy_extras/nodes_kandinsky5.py
View File

@ -13,7 +13,7 @@ class Kandinsky5ImageToVideo(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="Kandinsky5ImageToVideo",
category="model/conditioning/video_models",
category="model/conditioning/kandinsky",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -71,7 +71,7 @@ class NormalizeVideoLatentStart(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="NormalizeVideoLatentStart",
category="model/conditioning/video_models",
category="model/conditioning",
description="Normalizes the initial frames of a video latent to match the mean and standard deviation of subsequent reference frames. Helps reduce differences between the starting frames and the rest of the video.",
inputs=[
io.Latent.Input("latent"),
@ -104,8 +104,9 @@ class CLIPTextEncodeKandinsky5(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="CLIPTextEncodeKandinsky5",
display_name="CLIP Text Encode (Kandinsky 5)",
search_aliases=["kandinsky prompt"],
category="advanced/conditioning/kandinsky5",
category="model/conditioning/kandinsky",
inputs=[
io.Clip.Input("clip"),
io.String.Input("clip_l", multiline=True, dynamic_prompts=True),

2
comfy_extras/nodes_latent.py
View File

@ -262,6 +262,7 @@ class LatentBatch(io.ComfyNode):
return io.Schema(
node_id="LatentBatch",
search_aliases=["combine latents", "merge latents", "join latents"],
display_name="Batch Latents (DEPRECATED)",
category="model/latent/batch",
is_deprecated=True,
inputs=[
@ -447,6 +448,7 @@ class ReplaceVideoLatentFrames(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="ReplaceVideoLatentFrames",
display_name="Replace Video Latent Frames",
category="model/latent/batch",
inputs=[
io.Latent.Input("destination", tooltip="The destination latent where frames will be replaced."),

22
comfy_extras/nodes_lt.py
View File

@ -25,7 +25,7 @@ class GetICLoRAParameters(io.ComfyNode):
display_name="Get IC-LoRA Parameters",
description="Extracts IC-LoRA parameters from the safetensors metadata of a LoRA-loaded "
"model and outputs them for LTXVAddGuide (eg. reference_downscale_factor).",
category="model/conditioning/video_models",
category="model/conditioning/ltxv",
search_aliases=["ic-lora", "ic lora", "iclora", "downscale factor", "reference downscale"],
inputs=[
io.Model.Input(
@ -62,7 +62,7 @@ class EmptyLTXVLatentVideo(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="EmptyLTXVLatentVideo",
category="model/latent/video/ltxv",
category="model/latent/ltxv",
inputs=[
io.Int.Input("width", default=768, min=64, max=nodes.MAX_RESOLUTION, step=32),
io.Int.Input("height", default=512, min=64, max=nodes.MAX_RESOLUTION, step=32),
@ -86,7 +86,7 @@ class LTXVImgToVideo(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="LTXVImgToVideo",
category="model/conditioning/video_models",
category="model/conditioning/ltxv",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -131,7 +131,7 @@ class LTXVImgToVideoInplace(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="LTXVImgToVideoInplace",
category="model/conditioning/video_models",
category="model/conditioning/ltxv",
inputs=[
io.Vae.Input("vae"),
io.Image.Input("image"),
@ -251,7 +251,7 @@ class LTXVAddGuide(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="LTXVAddGuide",
category="model/conditioning/video_models",
category="model/conditioning/ltxv",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -498,7 +498,7 @@ class LTXVCropGuides(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="LTXVCropGuides",
category="model/conditioning/video_models",
category="model/conditioning/ltxv",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -542,7 +542,7 @@ class LTXVConditioning(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="LTXVConditioning",
category="model/conditioning/video_models",
category="model/conditioning/ltxv",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -566,7 +566,7 @@ class ModelSamplingLTXV(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="ModelSamplingLTXV",
category="advanced/model",
category="model/patch/ltxv",
inputs=[
io.Model.Input("model"),
io.Float.Input("max_shift", default=2.05, min=0.0, max=100.0, step=0.01),
@ -746,7 +746,7 @@ class LTXVConcatAVLatent(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="LTXVConcatAVLatent",
category="model/latent/video/ltxv",
category="model/latent/ltxv",
inputs=[
io.Latent.Input("video_latent"),
io.Latent.Input("audio_latent"),
@ -781,7 +781,7 @@ class LTXVSeparateAVLatent(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="LTXVSeparateAVLatent",
category="model/latent/video/ltxv",
category="model/latent/ltxv",
description="LTXV Separate AV Latent",
inputs=[
io.Latent.Input("av_latent"),
@ -814,7 +814,7 @@ class LTXVReferenceAudio(io.ComfyNode):
return io.Schema(
node_id="LTXVReferenceAudio",
display_name="LTXV Reference Audio (ID-LoRA)",
category="model/conditioning/audio",
category="model/conditioning/ltxv",
description="Set reference audio for ID-LoRA speaker identity transfer. Encodes a reference audio clip into the conditioning and optionally patches the model with identity guidance (extra forward pass without reference, amplifying the speaker identity effect).",
inputs=[
io.Model.Input("model"),

12
comfy_extras/nodes_lt_audio.py
View File

@ -40,7 +40,7 @@ class LTXVAudioVAEEncode(VAEEncodeAudio):
return io.Schema(
node_id="LTXVAudioVAEEncode",
display_name="LTXV Audio VAE Encode",
category="model/latent/audio",
category="model/latent/ltxv",
inputs=[
io.Audio.Input("audio", tooltip="The audio to be encoded."),
io.Vae.Input(
@ -63,7 +63,7 @@ class LTXVAudioVAEDecode(io.ComfyNode):
return io.Schema(
node_id="LTXVAudioVAEDecode",
display_name="LTXV Audio VAE Decode",
category="model/latent/audio",
category="model/latent/ltxv",
inputs=[
io.Latent.Input("samples", tooltip="The latent to be decoded."),
io.Vae.Input(
@ -96,7 +96,7 @@ class LTXVEmptyLatentAudio(io.ComfyNode):
return io.Schema(
node_id="LTXVEmptyLatentAudio",
display_name="LTXV Empty Latent Audio",
category="model/latent/audio",
category="model/latent/ltxv",
inputs=[
io.Int.Input(
"frames_number",
@ -168,9 +168,9 @@ class LTXAVTextEncoderLoader(io.ComfyNode):
def define_schema(cls) -> io.Schema:
return io.Schema(
node_id="LTXAVTextEncoderLoader",
display_name="LTXV Audio Text Encoder Loader",
category="advanced/loaders",
description="[Recipes]\n\nltxav: gemma 3 12B",
display_name="Load LTXV Audio Text Encoder",
category="model/loaders",
description="Recipes:\nltxav: gemma 3 12B",
inputs=[
io.Combo.Input(
"text_encoder",

2
comfy_extras/nodes_lt_upsampler.py
View File

@ -13,7 +13,7 @@ class LTXVLatentUpsampler(IO.ComfyNode):
def define_schema(cls):
return IO.Schema(
node_id="LTXVLatentUpsampler",
category="model/latent/video",
category="model/latent/ltxv",
is_experimental=True,
inputs=[
IO.Latent.Input("samples"),

6
comfy_extras/nodes_lumina2.py
View File

@ -9,7 +9,7 @@ class RenormCFG(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="RenormCFG",
category="advanced/model",
category="model/patch",
inputs=[
io.Model.Input("model"),
io.Float.Input("cfg_trunc", default=100, min=0.0, max=100.0, step=0.01, advanced=True),
@ -80,8 +80,8 @@ class CLIPTextEncodeLumina2(io.ComfyNode):
return io.Schema(
node_id="CLIPTextEncodeLumina2",
search_aliases=["lumina prompt"],
display_name="CLIP Text Encode for Lumina2",
category="model/conditioning",
display_name="CLIP Text Encode (Lumina 2)",
category="model/conditioning/lumina",
description="Encodes a system prompt and a user prompt using a CLIP model into an embedding "
"that can be used to guide the diffusion model towards generating specific images.",
inputs=[

1
comfy_extras/nodes_mask.py
View File

@ -53,6 +53,7 @@ class LatentCompositeMasked(IO.ComfyNode):
return IO.Schema(
node_id="LatentCompositeMasked",
search_aliases=["overlay latent", "layer latent", "paste latent", "inpaint latent"],
display_name="Latent Composite Masked",
category="model/latent",
inputs=[
IO.Latent.Input("destination"),

2
comfy_extras/nodes_mochi.py
View File

@ -10,7 +10,7 @@ class EmptyMochiLatentVideo(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="EmptyMochiLatentVideo",
category="model/latent/video",
category="model/latent/mochi",
inputs=[
io.Int.Input("width", default=848, min=16, max=nodes.MAX_RESOLUTION, step=16),
io.Int.Input("height", default=480, min=16, max=nodes.MAX_RESOLUTION, step=16),

19
comfy_extras/nodes_model_advanced.py
View File

@ -59,7 +59,7 @@ class ModelSamplingDiscrete:
RETURN_TYPES = ("MODEL",)
FUNCTION = "patch"
CATEGORY = "advanced/model"
CATEGORY = "model/patch"
def patch(self, model, sampling, zsnr):
m = model.clone()
@ -97,7 +97,7 @@ class ModelSamplingStableCascade:
RETURN_TYPES = ("MODEL",)
FUNCTION = "patch"
CATEGORY = "advanced/model"
CATEGORY = "model/patch/stable cascade"
def patch(self, model, shift):
m = model.clone()
@ -123,7 +123,7 @@ class ModelSamplingSD3:
RETURN_TYPES = ("MODEL",)
FUNCTION = "patch"
CATEGORY = "advanced/model"
CATEGORY = "model/patch/stable diffusion"
def patch(self, model, shift, multiplier=1000):
m = model.clone()
@ -150,6 +150,7 @@ class ModelSamplingAuraFlow(ModelSamplingSD3):
}}
FUNCTION = "patch_aura"
CATEGORY = "model/patch"
def patch_aura(self, model, shift):
return self.patch(model, shift, multiplier=1.0)
@ -167,7 +168,7 @@ class ModelSamplingFlux:
RETURN_TYPES = ("MODEL",)
FUNCTION = "patch"
CATEGORY = "advanced/model"
CATEGORY = "model/patch/flux"
def patch(self, model, max_shift, base_shift, width, height):
m = model.clone()
@ -202,7 +203,7 @@ class ModelSamplingContinuousEDM:
RETURN_TYPES = ("MODEL",)
FUNCTION = "patch"
CATEGORY = "advanced/model"
CATEGORY = "model/patch"
def patch(self, model, sampling, sigma_max, sigma_min):
m = model.clone()
@ -247,7 +248,7 @@ class ModelSamplingContinuousV:
RETURN_TYPES = ("MODEL",)
FUNCTION = "patch"
CATEGORY = "advanced/model"
CATEGORY = "model/patch"
def patch(self, model, sampling, sigma_max, sigma_min):
m = model.clone()
@ -273,7 +274,7 @@ class RescaleCFG:
RETURN_TYPES = ("MODEL",)
FUNCTION = "patch"
CATEGORY = "advanced/model"
CATEGORY = "model/patch"
def patch(self, model, multiplier):
def rescale_cfg(args):
@ -314,7 +315,7 @@ class ModelNoiseScale:
RETURN_TYPES = ("MODEL",)
FUNCTION = "patch"
CATEGORY = "advanced/model"
CATEGORY = "model/patch"
def patch(self, model, noise_scale):
m = model.clone()
@ -337,7 +338,7 @@ class ModelComputeDtype:
RETURN_TYPES = ("MODEL",)
FUNCTION = "patch"
CATEGORY = "advanced/debug/model"
CATEGORY = "advanced/debug"
def patch(self, model, dtype):
m = model.clone()

22
comfy_extras/nodes_model_merging.py
View File

@ -21,7 +21,7 @@ class ModelMergeSimple:
RETURN_TYPES = ("MODEL",)
FUNCTION = "merge"
CATEGORY = "advanced/model_merging"
CATEGORY = "model/merging"
def merge(self, model1, model2, ratio):
m = model1.clone()
@ -40,7 +40,7 @@ class ModelSubtract:
RETURN_TYPES = ("MODEL",)
FUNCTION = "merge"
CATEGORY = "advanced/model_merging"
CATEGORY = "model/merging"
def merge(self, model1, model2, multiplier):
m = model1.clone()
@ -58,7 +58,7 @@ class ModelAdd:
RETURN_TYPES = ("MODEL",)
FUNCTION = "merge"
CATEGORY = "advanced/model_merging"
CATEGORY = "model/merging"
def merge(self, model1, model2):
m = model1.clone()
@ -78,7 +78,7 @@ class CLIPMergeSimple:
RETURN_TYPES = ("CLIP",)
FUNCTION = "merge"
CATEGORY = "advanced/model_merging"
CATEGORY = "model/merging"
def merge(self, clip1, clip2, ratio):
m = clip1.clone()
@ -101,7 +101,7 @@ class CLIPSubtract:
RETURN_TYPES = ("CLIP",)
FUNCTION = "merge"
CATEGORY = "advanced/model_merging"
CATEGORY = "model/merging"
def merge(self, clip1, clip2, multiplier):
m = clip1.clone()
@ -123,7 +123,7 @@ class CLIPAdd:
RETURN_TYPES = ("CLIP",)
FUNCTION = "merge"
CATEGORY = "advanced/model_merging"
CATEGORY = "model/merging"
def merge(self, clip1, clip2):
m = clip1.clone()
@ -147,7 +147,7 @@ class ModelMergeBlocks:
RETURN_TYPES = ("MODEL",)
FUNCTION = "merge"
CATEGORY = "advanced/model_merging"
CATEGORY = "model/merging"
def merge(self, model1, model2, **kwargs):
m = model1.clone()
@ -242,7 +242,7 @@ class CheckpointSave:
FUNCTION = "save"
OUTPUT_NODE = True
CATEGORY = "advanced/model_merging"
CATEGORY = "model/merging"
def save(self, model, clip, vae, filename_prefix, prompt=None, extra_pnginfo=None):
save_checkpoint(model, clip=clip, vae=vae, filename_prefix=filename_prefix, output_dir=self.output_dir, prompt=prompt, extra_pnginfo=extra_pnginfo)
@ -261,7 +261,7 @@ class CLIPSave:
FUNCTION = "save"
OUTPUT_NODE = True
CATEGORY = "advanced/model_merging"
CATEGORY = "model/merging"
def save(self, clip, filename_prefix, prompt=None, extra_pnginfo=None):
prompt_info = ""
@ -318,7 +318,7 @@ class VAESave:
FUNCTION = "save"
OUTPUT_NODE = True
CATEGORY = "advanced/model_merging"
CATEGORY = "model/merging"
def save(self, vae, filename_prefix, prompt=None, extra_pnginfo=None):
full_output_folder, filename, counter, subfolder, filename_prefix = folder_paths.get_save_image_path(filename_prefix, self.output_dir)
@ -353,7 +353,7 @@ class ModelSave:
FUNCTION = "save"
OUTPUT_NODE = True
CATEGORY = "advanced/model_merging"
CATEGORY = "model/merging"
def save(self, model, filename_prefix, prompt=None, extra_pnginfo=None):
save_checkpoint(model, filename_prefix=filename_prefix, output_dir=self.output_dir, prompt=prompt, extra_pnginfo=extra_pnginfo)

28
comfy_extras/nodes_model_merging_model_specific.py
View File

@ -1,7 +1,7 @@
import comfy_extras.nodes_model_merging
class ModelMergeSD1(comfy_extras.nodes_model_merging.ModelMergeBlocks):
CATEGORY = "advanced/model_merging/model_specific"
CATEGORY = "model/merging/model specific"
@classmethod
def INPUT_TYPES(s):
arg_dict = { "model1": ("MODEL",),
@ -27,7 +27,7 @@ class ModelMergeSD1(comfy_extras.nodes_model_merging.ModelMergeBlocks):
class ModelMergeSDXL(comfy_extras.nodes_model_merging.ModelMergeBlocks):
CATEGORY = "advanced/model_merging/model_specific"
CATEGORY = "model/merging/model specific"
@classmethod
def INPUT_TYPES(s):
@ -53,7 +53,7 @@ class ModelMergeSDXL(comfy_extras.nodes_model_merging.ModelMergeBlocks):
return {"required": arg_dict}
class ModelMergeSD3_2B(comfy_extras.nodes_model_merging.ModelMergeBlocks):
CATEGORY = "advanced/model_merging/model_specific"
CATEGORY = "model/merging/model specific"
@classmethod
def INPUT_TYPES(s):
@ -77,7 +77,7 @@ class ModelMergeSD3_2B(comfy_extras.nodes_model_merging.ModelMergeBlocks):
class ModelMergeAuraflow(comfy_extras.nodes_model_merging.ModelMergeBlocks):
CATEGORY = "advanced/model_merging/model_specific"
CATEGORY = "model/merging/model specific"
@classmethod
def INPUT_TYPES(s):
@ -104,7 +104,7 @@ class ModelMergeAuraflow(comfy_extras.nodes_model_merging.ModelMergeBlocks):
return {"required": arg_dict}
class ModelMergeFlux1(comfy_extras.nodes_model_merging.ModelMergeBlocks):
CATEGORY = "advanced/model_merging/model_specific"
CATEGORY = "model/merging/model specific"
@classmethod
def INPUT_TYPES(s):
@ -130,7 +130,7 @@ class ModelMergeFlux1(comfy_extras.nodes_model_merging.ModelMergeBlocks):
return {"required": arg_dict}
class ModelMergeSD35_Large(comfy_extras.nodes_model_merging.ModelMergeBlocks):
CATEGORY = "advanced/model_merging/model_specific"
CATEGORY = "model/merging/model specific"
@classmethod
def INPUT_TYPES(s):
@ -153,7 +153,7 @@ class ModelMergeSD35_Large(comfy_extras.nodes_model_merging.ModelMergeBlocks):
return {"required": arg_dict}
class ModelMergeMochiPreview(comfy_extras.nodes_model_merging.ModelMergeBlocks):
CATEGORY = "advanced/model_merging/model_specific"
CATEGORY = "model/merging/model specific"
@classmethod
def INPUT_TYPES(s):
@ -175,7 +175,7 @@ class ModelMergeMochiPreview(comfy_extras.nodes_model_merging.ModelMergeBlocks):
return {"required": arg_dict}
class ModelMergeLTXV(comfy_extras.nodes_model_merging.ModelMergeBlocks):
CATEGORY = "advanced/model_merging/model_specific"
CATEGORY = "model/merging/model specific"
@classmethod
def INPUT_TYPES(s):
@ -197,7 +197,7 @@ class ModelMergeLTXV(comfy_extras.nodes_model_merging.ModelMergeBlocks):
return {"required": arg_dict}
class ModelMergeCosmos7B(comfy_extras.nodes_model_merging.ModelMergeBlocks):
CATEGORY = "advanced/model_merging/model_specific"
CATEGORY = "model/merging/model specific"
@classmethod
def INPUT_TYPES(s):
@ -221,7 +221,7 @@ class ModelMergeCosmos7B(comfy_extras.nodes_model_merging.ModelMergeBlocks):
return {"required": arg_dict}
class ModelMergeCosmos14B(comfy_extras.nodes_model_merging.ModelMergeBlocks):
CATEGORY = "advanced/model_merging/model_specific"
CATEGORY = "model/merging/model specific"
@classmethod
def INPUT_TYPES(s):
@ -245,7 +245,7 @@ class ModelMergeCosmos14B(comfy_extras.nodes_model_merging.ModelMergeBlocks):
return {"required": arg_dict}
class ModelMergeWAN2_1(comfy_extras.nodes_model_merging.ModelMergeBlocks):
CATEGORY = "advanced/model_merging/model_specific"
CATEGORY = "model/merging/model specific"
DESCRIPTION = "1.3B model has 30 blocks, 14B model has 40 blocks. Image to video model has the extra img_emb."
@classmethod
@ -269,7 +269,7 @@ class ModelMergeWAN2_1(comfy_extras.nodes_model_merging.ModelMergeBlocks):
return {"required": arg_dict}
class ModelMergeCosmosPredict2_2B(comfy_extras.nodes_model_merging.ModelMergeBlocks):
CATEGORY = "advanced/model_merging/model_specific"
CATEGORY = "model/merging/model specific"
@classmethod
def INPUT_TYPES(s):
@ -292,7 +292,7 @@ class ModelMergeCosmosPredict2_2B(comfy_extras.nodes_model_merging.ModelMergeBlo
return {"required": arg_dict}
class ModelMergeCosmosPredict2_14B(comfy_extras.nodes_model_merging.ModelMergeBlocks):
CATEGORY = "advanced/model_merging/model_specific"
CATEGORY = "model/merging/model specific"
@classmethod
def INPUT_TYPES(s):
@ -315,7 +315,7 @@ class ModelMergeCosmosPredict2_14B(comfy_extras.nodes_model_merging.ModelMergeBl
return {"required": arg_dict}
class ModelMergeQwenImage(comfy_extras.nodes_model_merging.ModelMergeBlocks):
CATEGORY = "advanced/model_merging/model_specific"
CATEGORY = "model/merging/model specific"
@classmethod
def INPUT_TYPES(s):

14
comfy_extras/nodes_model_patch.py
View File

@ -232,7 +232,7 @@ class ModelPatchLoader:
FUNCTION = "load_model_patch"
EXPERIMENTAL = True
CATEGORY = "advanced/loaders"
CATEGORY = "model/loaders"
def load_model_patch(self, name):
model_patch_path = folder_paths.get_full_path_or_raise("model_patches", name)
@ -479,7 +479,7 @@ class QwenImageDiffsynthControlnet:
FUNCTION = "diffsynth_controlnet"
EXPERIMENTAL = True
CATEGORY = "advanced/loaders/qwen"
CATEGORY = "model/patch/qwen"
def diffsynth_controlnet(self, model, model_patch, vae, image=None, strength=1.0, inpaint_image=None, mask=None):
model_patched = model.clone()
@ -512,7 +512,7 @@ class ZImageFunControlnet(QwenImageDiffsynthControlnet):
},
"optional": {"image": ("IMAGE",), "inpaint_image": ("IMAGE",), "mask": ("MASK",)}}
CATEGORY = "advanced/loaders/zimage"
CATEGORY = "model/patch/z-image"
class UsoStyleProjectorPatch:
def __init__(self, model_patch, encoded_image):
@ -675,3 +675,11 @@ NODE_CLASS_MAPPINGS = {
"USOStyleReference": USOStyleReference,
"SUPIRApply": SUPIRApply,
}
NODE_DISPLAY_NAME_MAPPINGS = {
"ModelPatchLoader": "Load Model Patch",
"QwenImageDiffsynthControlnet": "Apply Qwen Image DiffSynth ControlNet",
"ZImageFunControlnet": "Apply Z-Image Fun ControlNet",
"USOStyleReference": "Apply USO Style Reference",
"SUPIRApply": "Apply SUPIR Patch",
}

6
comfy_extras/nodes_pid.py
View File

@ -14,10 +14,8 @@ class PiDConditioning(io.ComfyNode):
return io.Schema(
node_id="PiDConditioning",
display_name="PiD Conditioning",
category="advanced/conditioning",
description=(
"Attaches a latent and a degrade_sigma scalar to a CONDITIONING for PiD decoding/upscaling"
),
category="model/conditioning",
description=("Attaches a latent and a degrade_sigma scalar to a CONDITIONING for PiD decoding/upscaling"),
inputs=[
io.Conditioning.Input("positive"),
io.Latent.Input("latent", tooltip="latent (from VAEEncode or a KSampler)."),

3
comfy_extras/nodes_pixart.py
View File

@ -7,8 +7,9 @@ class CLIPTextEncodePixArtAlpha(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="CLIPTextEncodePixArtAlpha",
display_name="CLIP Text Encode (PixArt Alpha)",
search_aliases=["pixart prompt"],
category="advanced/conditioning",
category="model/conditioning/pixart",
description="Encodes text and sets the resolution conditioning for PixArt Alpha. Does not apply to PixArt Sigma.",
inputs=[
io.Int.Input("width", default=1024, min=0, max=nodes.MAX_RESOLUTION),

2
comfy_extras/nodes_post_processing.py
View File

@ -616,7 +616,7 @@ class BatchLatentsNode(io.ComfyNode):
node_id="BatchLatentsNode",
search_aliases=["combine latents", "stack latents", "merge latents"],
display_name="Batch Latents",
category="model/latent",
category="model/latent/batch",
inputs=[
io.Autogrow.Input("latents", template=autogrow_template)
],

4
comfy_extras/nodes_qwen.py
View File

@ -12,7 +12,7 @@ class TextEncodeQwenImageEdit(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="TextEncodeQwenImageEdit",
category="advanced/conditioning",
category="model/conditioning/qwen image",
inputs=[
io.Clip.Input("clip"),
io.String.Input("prompt", multiline=True, dynamic_prompts=True),
@ -55,7 +55,7 @@ class TextEncodeQwenImageEditPlus(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="TextEncodeQwenImageEditPlus",
category="advanced/conditioning",
category="model/conditioning/qwen image",
inputs=[
io.Clip.Input("clip"),
io.String.Input("prompt", multiline=True, dynamic_prompts=True),

16
comfy_extras/nodes_scail.py
View File

@ -123,7 +123,7 @@ class WanSCAILToVideo(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="WanSCAILToVideo",
category="model/conditioning/video_models",
category="model/conditioning/wan/scail",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -257,18 +257,16 @@ class SCAIL2ColoredMask(io.ComfyNode):
return io.Schema(
node_id="SCAIL2ColoredMask",
display_name="Create SCAIL-2 Colored Mask",
category="conditioning/video_models/scail",
category="model/conditioning/wan/scail",
inputs=[
SAM3TrackData.Input("driving_track_data", tooltip="SAM3 track of the driving pose video. Will be rendered into the pose_video_mask output."),
SAM3TrackData.Input("ref_track_data", optional=True,
tooltip="SAM3 track of the reference image."),
io.String.Input("object_indices", default="",
tooltip="Comma-separated list of person indices to include (e.g. '0,2,3'). Applied to both reference and pose video masks. Empty = all."),
SAM3TrackData.Input("ref_track_data", optional=True, tooltip="SAM3 track of the reference image."),
io.String.Input("object_indices", default="", tooltip="Comma-separated list of person indices to include (e.g. '0,2,3'). Applied to both reference and pose video masks. Empty = all."),
io.Combo.Input("sort_by", options=["none", "left_to_right", "area"], default="left_to_right",
tooltip="Order in which palette colors are assigned to the tracked objects (applied to both reference and pose video so each identity keeps the same color). left_to_right = leftmost object (by first-frame centroid) gets the first color; area = biggest object (by first-frame mask area) gets the first color; none = keep SAM3's order."),
tooltip="Order in which palette colors are assigned to the tracked objects (applied to both reference and pose video so each identity keeps the same color). left_to_right = leftmost object (by first-frame centroid) gets the first color; area = biggest object (by first-frame mask area) gets the first color; none = keep SAM3's order."),
io.Boolean.Input("replacement_mode", default=False,
tooltip="False = Animation Mode (pose_video_mask has black background, reference_image_mask has white background). "
"True = Replacement Mode (pose_video_mask has white background, reference_image_mask has black background)."),
tooltip="False = Animation Mode (pose_video_mask has black background, reference_image_mask has white background). "
"True = Replacement Mode (pose_video_mask has white background, reference_image_mask has black background)."),
],
outputs=[
io.Image.Output("pose_video_mask"),

10
comfy_extras/nodes_sd3.py
View File

@ -13,8 +13,9 @@ class TripleCLIPLoader(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="TripleCLIPLoader",
category="advanced/loaders",
description="[Recipes]\n\nsd3: clip-l, clip-g, t5",
display_name="Load CLIP (Triple)",
category="model/loaders",
description="Recipes:\nsd3: clip-l, clip-g, t5",
inputs=[
io.Combo.Input("clip_name1", options=folder_paths.get_filename_list("text_encoders")),
io.Combo.Input("clip_name2", options=folder_paths.get_filename_list("text_encoders")),
@ -41,7 +42,7 @@ class EmptySD3LatentImage(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="EmptySD3LatentImage",
category="model/latent/sd3",
category="model/latent/stable diffusion",
inputs=[
io.Int.Input("width", default=1024, min=16, max=nodes.MAX_RESOLUTION, step=16),
io.Int.Input("height", default=1024, min=16, max=nodes.MAX_RESOLUTION, step=16),
@ -66,7 +67,8 @@ class CLIPTextEncodeSD3(io.ComfyNode):
return io.Schema(
node_id="CLIPTextEncodeSD3",
search_aliases=["sd3 prompt"],
category="advanced/conditioning",
display_name="CLIP Text Encode (SD3)",
category="model/conditioning/stable diffusion",
inputs=[
io.Clip.Input("clip"),
io.String.Input("clip_l", multiline=True, dynamic_prompts=True),

41
comfy_extras/nodes_sdpose.py
View File

@ -96,8 +96,12 @@ class KeypointDraw:
# Body connections - matching DWPose limbSeq (1-indexed, converted to 0-indexed)
self.body_limbSeq = [
[2, 3], [2, 6], [3, 4], [4, 5], [6, 7], [7, 8], [2, 9], [9, 10],
[10, 11], [2, 12], [12, 13], [13, 14], [2, 1], [1, 15], [15, 17],
[1, 16], [16, 18]
[10, 11], [2, 12], [12, 13], [13, 14]
]
# Head connections (1-indexed, converted to 0-indexed)
self.head_edges = [
[2, 1], [1, 15], [15, 17], [1, 16], [16, 18]
]
# Colors matching DWPose
@ -215,7 +219,7 @@ class KeypointDraw:
return unique_pts if len(unique_pts) > 1 else [[center[0], center[1]], [center[0], center[1]]]
def draw_wholebody_keypoints(self, canvas, keypoints, scores=None, threshold=0.3,
draw_body=True, draw_feet=True, draw_face=True, draw_hands=True, stick_width=4, face_point_size=3):
draw_body=True, draw_head=True, draw_feet=True, draw_face=True, draw_hands=True, stick_width=4, face_point_size=3):
"""
Draw wholebody keypoints (134 keypoints after processing) in DWPose style.
@ -237,9 +241,17 @@ class KeypointDraw:
"""
H, W, C = canvas.shape
# Draw body limbs
if draw_body and len(keypoints) >= 18:
for i, limb in enumerate(self.body_limbSeq):
# Draw body limbs & head connections
if (draw_body or draw_head) and len(keypoints) >= 18:
colorIndexOffset = 0
edges = []
if draw_body:
edges += self.body_limbSeq
else:
colorIndexOffset += len(self.body_limbSeq)
if draw_head:
edges += self.head_edges
for i, limb in enumerate(edges):
# Convert from 1-indexed to 0-indexed
idx1, idx2 = limb[0] - 1, limb[1] - 1
@ -262,11 +274,17 @@ class KeypointDraw:
polygon = self.draw.ellipse2Poly((int(mY), int(mX)), (int(length / 2), stick_width), int(angle), 0, 360, 1)
self.draw.fillConvexPoly(canvas, polygon, self.colors[i % len(self.colors)])
self.draw.fillConvexPoly(canvas, polygon, self.colors[(i + colorIndexOffset) % len(self.colors)])
# Draw body keypoints
if draw_body and len(keypoints) >= 18:
# Draw body & head keypoints
if (draw_body or draw_head) and len(keypoints) >= 18:
head_keypoints = {0, 14, 15, 16, 17} # nose, eyes, ears
neck_point = 1
for i in range(18):
if not draw_head and i in head_keypoints:
continue
if not draw_body and i not in head_keypoints and i != neck_point:
continue
if scores is not None and scores[i] < threshold:
continue
x, y = int(keypoints[i][0]), int(keypoints[i][1])
@ -365,6 +383,7 @@ class SDPoseDrawKeypoints(io.ComfyNode):
io.Int.Input("stick_width", default=4, min=1, max=10, step=1),
io.Int.Input("face_point_size", default=3, min=1, max=10, step=1),
io.Float.Input("score_threshold", default=0.3, min=0.0, max=1.0, step=0.01),
io.Boolean.Input("draw_head", default=True),
],
outputs=[
io.Image.Output(),
@ -372,7 +391,7 @@ class SDPoseDrawKeypoints(io.ComfyNode):
)
@classmethod
def execute(cls, keypoints, draw_body, draw_hands, draw_face, draw_feet, stick_width, face_point_size, score_threshold) -> io.NodeOutput:
def execute(cls, keypoints, draw_body, draw_hands, draw_face, draw_feet, stick_width, face_point_size, score_threshold, draw_head) -> io.NodeOutput:
if not keypoints:
return io.NodeOutput(torch.zeros((1, 64, 64, 3), dtype=torch.float32))
height = keypoints[0]["canvas_height"]
@ -405,7 +424,7 @@ class SDPoseDrawKeypoints(io.ComfyNode):
canvas = drawer.draw_wholebody_keypoints(
canvas, kp, sc,
threshold=score_threshold,
draw_body=draw_body, draw_feet=draw_feet,
draw_body=draw_body, draw_head=draw_head, draw_feet=draw_feet,
draw_face=draw_face, draw_hands=draw_hands,
stick_width=stick_width, face_point_size=face_point_size,
)

2
comfy_extras/nodes_sdupscale.py
View File

@ -9,7 +9,7 @@ class SD_4XUpscale_Conditioning(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="SD_4XUpscale_Conditioning",
category="model/conditioning/upscale_diffusion",
category="model/conditioning/stable diffusion upscaler",
inputs=[
io.Image.Input("images"),
io.Conditioning.Input("positive"),

6
comfy_extras/nodes_stable3d.py
View File

@ -27,7 +27,7 @@ class StableZero123_Conditioning(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="StableZero123_Conditioning",
category="model/conditioning/3d_models",
category="model/conditioning/stable zero123",
inputs=[
io.ClipVision.Input("clip_vision"),
io.Image.Input("init_image"),
@ -65,7 +65,7 @@ class StableZero123_Conditioning_Batched(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="StableZero123_Conditioning_Batched",
category="model/conditioning/3d_models",
category="model/conditioning/stable zero123",
inputs=[
io.ClipVision.Input("clip_vision"),
io.Image.Input("init_image"),
@ -112,7 +112,7 @@ class SV3D_Conditioning(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="SV3D_Conditioning",
category="model/conditioning/3d_models",
category="model/conditioning/stable video 3d",
inputs=[
io.ClipVision.Input("clip_vision"),
io.Image.Input("init_image"),

6
comfy_extras/nodes_stable_cascade.py
View File

@ -29,7 +29,7 @@ class StableCascade_EmptyLatentImage(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="StableCascade_EmptyLatentImage",
category="model/latent/stable_cascade",
category="model/latent/stable cascade",
inputs=[
io.Int.Input("width", default=1024, min=256, max=nodes.MAX_RESOLUTION, step=8),
io.Int.Input("height", default=1024, min=256, max=nodes.MAX_RESOLUTION, step=8),
@ -58,7 +58,7 @@ class StableCascade_StageC_VAEEncode(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="StableCascade_StageC_VAEEncode",
category="model/latent/stable_cascade",
category="model/latent/stable cascade",
inputs=[
io.Image.Input("image"),
io.Vae.Input("vae"),
@ -93,7 +93,7 @@ class StableCascade_StageB_Conditioning(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="StableCascade_StageB_Conditioning",
category="model/conditioning/stable_cascade",
category="model/conditioning/stable cascade",
inputs=[
io.Conditioning.Input("conditioning"),
io.Latent.Input("stage_c"),

2
comfy_extras/nodes_textgen.py
View File

@ -35,7 +35,7 @@ class TextGenerate(io.ComfyNode):
io.Image.Input("image", optional=True),
io.Image.Input("video", optional=True, tooltip="Video frames as image batch. Assumed to be 24 FPS; subsampled to 1 FPS internally."),
io.Audio.Input("audio", optional=True),
io.Int.Input("max_length", default=256, min=1, max=2048),
io.Int.Input("max_length", default=512, min=1, max=32768),
io.DynamicCombo.Input("sampling_mode", options=sampling_options, display_name="Sampling Mode"),
io.Boolean.Input("thinking", optional=True, default=False, tooltip="Operate in thinking mode if the model supports it."),
io.Boolean.Input("use_default_template", optional=True, default=True, tooltip="Use the built in system prompt/template if the model has one.", advanced=True),

2
comfy_extras/nodes_train.py
View File

@ -1367,7 +1367,7 @@ class SaveLoRA(io.ComfyNode):
node_id="SaveLoRA",
search_aliases=["export lora"],
display_name="Save LoRA Weights",
category="advanced/model_merging",
category="model/merging",
is_experimental=True,
is_output_node=True,
inputs=[

7
comfy_extras/nodes_video_model.py
View File

@ -41,7 +41,7 @@ class SVD_img2vid_Conditioning:
FUNCTION = "encode"
CATEGORY = "model/conditioning/video_models"
CATEGORY = "model/conditioning/stable video"
def encode(self, clip_vision, init_image, vae, width, height, video_frames, motion_bucket_id, fps, augmentation_level):
output = clip_vision.encode_image(init_image)
@ -108,7 +108,7 @@ class VideoTriangleCFGGuidance:
return (m, )
class ImageOnlyCheckpointSave(comfy_extras.nodes_model_merging.CheckpointSave):
CATEGORY = "advanced/model_merging"
CATEGORY = "model/merging"
@classmethod
def INPUT_TYPES(s):
@ -138,7 +138,7 @@ class ConditioningSetAreaPercentageVideo:
RETURN_TYPES = ("CONDITIONING",)
FUNCTION = "append"
CATEGORY = "model/conditioning"
CATEGORY = "model/conditioning/transform"
def append(self, conditioning, width, height, temporal, x, y, z, strength):
c = node_helpers.conditioning_set_values(conditioning, {"area": ("percentage", temporal, height, width, z, y, x),
@ -160,4 +160,5 @@ NODE_DISPLAY_NAME_MAPPINGS = {
"ImageOnlyCheckpointLoader": "Load Checkpoint Image Only (img2vid model)",
"VideoLinearCFGGuidance": "Video Linear CFG Guidance",
"VideoTriangleCFGGuidance": "Video Triangle CFG Guidance",
"ConditioningSetAreaPercentageVideo": "Conditioning (Set Area with Percentage for Video)",
}

6
comfy_extras/nodes_void.py
View File

@ -175,7 +175,7 @@ class VOIDInpaintConditioning(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="VOIDInpaintConditioning",
category="model/conditioning/video_models",
category="model/conditioning/void",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -288,7 +288,7 @@ class VOIDWarpedNoise(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="VOIDWarpedNoise",
category="model/latent/video",
category="model/latent/void",
inputs=[
OpticalFlow.Input(
"optical_flow",
@ -393,7 +393,7 @@ class VOIDWarpedNoiseSource(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="VOIDWarpedNoiseSource",
category="model/sampling/noise",
category="model/latent/void",
inputs=[
io.Latent.Input("warped_noise",
tooltip="Warped noise latent from VOIDWarpedNoise"),

33
comfy_extras/nodes_wan.py
View File

@ -18,7 +18,7 @@ class WanImageToVideo(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="WanImageToVideo",
category="model/conditioning/video_models",
category="model/conditioning/wan",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -66,7 +66,7 @@ class WanFunControlToVideo(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="WanFunControlToVideo",
category="model/conditioning/video_models",
category="model/conditioning/wan/fun control",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -119,7 +119,7 @@ class Wan22FunControlToVideo(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="Wan22FunControlToVideo",
category="model/conditioning/video_models",
category="model/conditioning/wan/fun control",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -184,7 +184,7 @@ class WanFirstLastFrameToVideo(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="WanFirstLastFrameToVideo",
category="model/conditioning/video_models",
category="model/conditioning/wan",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -256,7 +256,7 @@ class WanFunInpaintToVideo(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="WanFunInpaintToVideo",
category="model/conditioning/video_models",
category="model/conditioning/wan/fun inpaint",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -288,7 +288,7 @@ class WanVaceToVideo(io.ComfyNode):
return io.Schema(
node_id="WanVaceToVideo",
search_aliases=["video conditioning", "video control"],
category="model/conditioning/video_models",
category="model/conditioning/wan/vace",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -375,7 +375,8 @@ class TrimVideoLatent(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="TrimVideoLatent",
category="model/latent/video",
display_name="Trim Video Latent",
category="model/latent",
inputs=[
io.Latent.Input("samples"),
io.Int.Input("trim_amount", default=0, min=0, max=99999),
@ -398,7 +399,7 @@ class WanCameraImageToVideo(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="WanCameraImageToVideo",
category="model/conditioning/video_models",
category="model/conditioning/wan/camera",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -452,7 +453,7 @@ class WanPhantomSubjectToVideo(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="WanPhantomSubjectToVideo",
category="model/conditioning/video_models",
category="model/conditioning/wan/phantom subject",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -707,7 +708,7 @@ class WanTrackToVideo(io.ComfyNode):
return io.Schema(
node_id="WanTrackToVideo",
search_aliases=["motion tracking", "trajectory video", "point tracking", "keypoint animation"],
category="model/conditioning/video_models",
category="model/conditioning/wan/move",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -951,7 +952,7 @@ class WanSoundImageToVideo(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="WanSoundImageToVideo",
category="model/conditioning/video_models",
category="model/conditioning/wan/sound",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -984,7 +985,7 @@ class WanSoundImageToVideoExtend(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="WanSoundImageToVideoExtend",
category="model/conditioning/video_models",
category="model/conditioning/wan/sound",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -1046,7 +1047,7 @@ class WanHuMoImageToVideo(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="WanHuMoImageToVideo",
category="model/conditioning/video_models",
category="model/conditioning/wan/humo",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -1112,7 +1113,7 @@ class WanAnimateToVideo(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="WanAnimateToVideo",
category="model/conditioning/video_models",
category="model/conditioning/wan/animate",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),
@ -1252,7 +1253,7 @@ class Wan22ImageToVideoLatent(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="Wan22ImageToVideoLatent",
category="model/conditioning/inpaint",
category="model/conditioning/wan",
inputs=[
io.Vae.Input("vae"),
io.Int.Input("width", default=1280, min=32, max=nodes.MAX_RESOLUTION, step=32),
@ -1302,7 +1303,7 @@ class WanInfiniteTalkToVideo(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="WanInfiniteTalkToVideo",
category="model/conditioning/video_models",
category="model/conditioning/wan/infinite talk",
inputs=[
io.DynamicCombo.Input("mode", options=[
io.DynamicCombo.Option("single_speaker", []),

4
comfy_extras/nodes_wandancer.py
View File

@ -713,7 +713,7 @@ class WanDancerEncodeAudio(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="WanDancerEncodeAudio",
category="model/conditioning/video_models",
category="model/conditioning/wan/dancer",
inputs=[
io.Audio.Input("audio"),
io.Int.Input("video_frames", default=149, min=1, max=nodes.MAX_RESOLUTION, step=4),
@ -787,7 +787,7 @@ class WanDancerVideo(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="WanDancerVideo",
category="model/conditioning/video_models",
category="model/conditioning/wan/dancer",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),

11
comfy_extras/nodes_wanmove.py
View File

@ -247,7 +247,7 @@ class WanMoveVisualizeTracks(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="WanMoveVisualizeTracks",
category="model/conditioning/video_models",
category="model/conditioning/wan/move",
inputs=[
io.Image.Input("images"),
io.Tracks.Input("tracks", optional=True),
@ -283,7 +283,7 @@ class WanMoveTracksFromCoords(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="WanMoveTracksFromCoords",
category="model/conditioning/video_models",
category="model/conditioning/wan/move",
inputs=[
io.String.Input("track_coords", force_input=True, default="[]", optional=True),
io.Mask.Input("track_mask", optional=True),
@ -325,7 +325,8 @@ class GenerateTracks(io.ComfyNode):
return io.Schema(
node_id="GenerateTracks",
search_aliases=["motion paths", "camera movement", "trajectory"],
category="model/conditioning/video_models",
display_name="Generate Video Tracks",
category="model/conditioning/wan/move",
inputs=[
io.Int.Input("width", default=832, min=16, max=4096, step=16),
io.Int.Input("height", default=480, min=16, max=4096, step=16),
@ -434,7 +435,7 @@ class WanMoveConcatTrack(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="WanMoveConcatTrack",
category="model/conditioning/video_models",
category="model/conditioning/wan/move",
inputs=[
io.Tracks.Input("tracks_1"),
io.Tracks.Input("tracks_2", optional=True),
@ -463,7 +464,7 @@ class WanMoveTrackToVideo(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="WanMoveTrackToVideo",
category="model/conditioning/video_models",
category="model/conditioning/wan/move",
inputs=[
io.Conditioning.Input("positive"),
io.Conditioning.Input("negative"),

2
comfy_extras/nodes_zimage.py
View File

@ -10,7 +10,7 @@ class TextEncodeZImageOmni(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="TextEncodeZImageOmni",
category="advanced/conditioning",
category="model/conditioning/z-image",
is_experimental=True,
inputs=[
io.Clip.Input("clip"),

4
main.py
View File

@ -127,6 +127,10 @@ def apply_custom_paths():
for config_path in itertools.chain(*args.extra_model_paths_config):
utils.extra_config.load_extra_path_config(config_path)
# --base-directory
if args.base_directory:
logging.info(f"Setting base directory to: {folder_paths.base_path}")
# --output-directory, --input-directory, --user-directory
if args.output_directory:
output_dir = os.path.abspath(args.output_directory)

58
nodes.py
View File

@ -87,7 +87,7 @@ class ConditioningCombine:
RETURN_TYPES = ("CONDITIONING",)
FUNCTION = "combine"
CATEGORY = "model/conditioning"
CATEGORY = "model/conditioning/transform"
SEARCH_ALIASES = ["combine", "merge conditioning", "combine prompts", "merge prompts", "mix prompts", "add prompt"]
def combine(self, conditioning_1, conditioning_2):
@ -104,7 +104,7 @@ class ConditioningAverage :
RETURN_TYPES = ("CONDITIONING",)
FUNCTION = "addWeighted"
CATEGORY = "model/conditioning"
CATEGORY = "model/conditioning/transform"
def addWeighted(self, conditioning_to, conditioning_from, conditioning_to_strength):
out = []
@ -143,7 +143,7 @@ class ConditioningConcat:
RETURN_TYPES = ("CONDITIONING",)
FUNCTION = "concat"
CATEGORY = "model/conditioning"
CATEGORY = "model/conditioning/transform"
def concat(self, conditioning_to, conditioning_from):
out = []
@ -176,7 +176,7 @@ class ConditioningSetArea:
RETURN_TYPES = ("CONDITIONING",)
FUNCTION = "append"
CATEGORY = "model/conditioning"
CATEGORY = "model/conditioning/transform"
def append(self, conditioning, width, height, x, y, strength):
c = node_helpers.conditioning_set_values(conditioning, {"area": (height // 8, width // 8, y // 8, x // 8),
@ -197,7 +197,7 @@ class ConditioningSetAreaPercentage:
RETURN_TYPES = ("CONDITIONING",)
FUNCTION = "append"
CATEGORY = "model/conditioning"
CATEGORY = "model/conditioning/transform"
def append(self, conditioning, width, height, x, y, strength):
c = node_helpers.conditioning_set_values(conditioning, {"area": ("percentage", height, width, y, x),
@ -214,7 +214,7 @@ class ConditioningSetAreaStrength:
RETURN_TYPES = ("CONDITIONING",)
FUNCTION = "append"
CATEGORY = "model/conditioning"
CATEGORY = "model/conditioning/transform"
def append(self, conditioning, strength):
c = node_helpers.conditioning_set_values(conditioning, {"strength": strength})
@ -234,7 +234,7 @@ class ConditioningSetMask:
RETURN_TYPES = ("CONDITIONING",)
FUNCTION = "append"
CATEGORY = "model/conditioning"
CATEGORY = "model/conditioning/transform"
def append(self, conditioning, mask, set_cond_area, strength):
set_area_to_bounds = False
@ -257,7 +257,7 @@ class ConditioningZeroOut:
RETURN_TYPES = ("CONDITIONING",)
FUNCTION = "zero_out"
CATEGORY = "advanced/conditioning"
CATEGORY = "model/conditioning/transform"
def zero_out(self, conditioning):
c = []
@ -283,11 +283,10 @@ class ConditioningSetTimestepRange:
RETURN_TYPES = ("CONDITIONING",)
FUNCTION = "set_range"
CATEGORY = "advanced/conditioning"
CATEGORY = "model/conditioning/transform"
def set_range(self, conditioning, start, end):
c = node_helpers.conditioning_set_values(conditioning, {"start_percent": start,
"end_percent": end})
c = node_helpers.conditioning_set_values(conditioning, {"start_percent": start, "end_percent": end})
return (c, )
class VAEDecode:
@ -389,7 +388,7 @@ class VAEEncodeForInpaint:
RETURN_TYPES = ("LATENT",)
FUNCTION = "encode"
CATEGORY = "model/latent/inpaint"
CATEGORY = "model/latent"
def encode(self, vae, pixels, mask, grow_mask_by=6):
downscale_ratio = vae.spacial_compression_encode()
@ -438,7 +437,7 @@ class InpaintModelConditioning:
RETURN_NAMES = ("positive", "negative", "latent")
FUNCTION = "encode"
CATEGORY = "model/conditioning/inpaint"
CATEGORY = "model/conditioning"
def encode(self, positive, negative, pixels, vae, mask, noise_mask=True):
x = (pixels.shape[1] // 8) * 8
@ -579,7 +578,7 @@ class CheckpointLoader:
RETURN_TYPES = ("MODEL", "CLIP", "VAE")
FUNCTION = "load_checkpoint"
CATEGORY = "advanced/loaders"
CATEGORY = "model/loaders"
DEPRECATED = True
def load_checkpoint(self, config_name, ckpt_name):
@ -625,8 +624,9 @@ class DiffusersLoader:
return {"required": {"model_path": (paths,), }}
RETURN_TYPES = ("MODEL", "CLIP", "VAE")
FUNCTION = "load_checkpoint"
DEPRECATED = True
CATEGORY = "advanced/loaders/deprecated"
CATEGORY = "model/loaders"
def load_checkpoint(self, model_path, output_vae=True, output_clip=True):
for search_path in folder_paths.get_folder_paths("diffusers"):
@ -952,7 +952,7 @@ class UNETLoader:
RETURN_TYPES = ("MODEL",)
FUNCTION = "load_unet"
CATEGORY = "advanced/loaders"
CATEGORY = "model/loaders"
def load_unet(self, unet_name, weight_dtype):
model_options = {}
@ -980,9 +980,9 @@ class CLIPLoader:
RETURN_TYPES = ("CLIP",)
FUNCTION = "load_clip"
CATEGORY = "advanced/loaders"
CATEGORY = "model/loaders"
DESCRIPTION = "[Recipes]\n\nstable_diffusion: clip-l\nstable_cascade: clip-g\nsd3: t5 xxl/ clip-g / clip-l\nstable_audio: t5 base\nmochi: t5 xxl\ncogvideox: t5 xxl (226-token padding)\ncosmos: old t5 xxl\nlumina2: gemma 2 2B\nwan: umt5 xxl\n hidream: llama-3.1 (Recommend) or t5\nomnigen2: qwen vl 2.5 3B\nlens: gpt-oss-20b\n pixeldit: gemma 2 2B elm"
DESCRIPTION = "Recipes:\nsd: clip-l\nstable cascade: clip-g\nsd3: t5 xxl / clip-g / clip-l\nstable audio: t5 base\nmochi: t5 xxl\ncogvideox: t5 xxl (226-token padding)\ncosmos: old t5 xxl\nlumina2: gemma 2 2B\nwan: umt5 xxl\nhidream: llama-3.1 (Recommend) or t5\nomnigen2: qwen vl 2.5 3B\nlens: gpt-oss-20b\npixeldit: gemma 2 2B elm"
def load_clip(self, clip_name, type="stable_diffusion", device="default"):
clip_type = getattr(comfy.sd.CLIPType, type.upper(), comfy.sd.CLIPType.STABLE_DIFFUSION)
@ -1008,9 +1008,9 @@ class DualCLIPLoader:
RETURN_TYPES = ("CLIP",)
FUNCTION = "load_clip"
CATEGORY = "advanced/loaders"
CATEGORY = "model/loaders"
DESCRIPTION = "[Recipes]\n\nsdxl: clip-l, clip-g\nsd3: clip-l, clip-g / clip-l, t5 / clip-g, t5\nflux: clip-l, t5\nhidream: at least one of t5 or llama, recommended t5 and llama\nhunyuan_image: qwen2.5vl 7b and byt5 small\nnewbie: gemma-3-4b-it, jina clip v2"
DESCRIPTION = "Recipes:\nsdxl: clip-l, clip-g\nsd3: clip-l, clip-g / clip-l, t5 / clip-g, t5\nflux: clip-l, t5\nhidream: at least one of t5 or llama, recommended t5 and llama\nhunyuan_image: qwen2.5vl 7b and byt5 small\nnewbie: gemma-3-4b-it, jina clip v2"
def load_clip(self, clip_name1, clip_name2, type, device="default"):
clip_type = getattr(comfy.sd.CLIPType, type.upper(), comfy.sd.CLIPType.STABLE_DIFFUSION)
@ -1091,7 +1091,7 @@ class StyleModelApply:
RETURN_TYPES = ("CONDITIONING",)
FUNCTION = "apply_stylemodel"
CATEGORY = "model/conditioning/style_model"
CATEGORY = "model/conditioning"
def apply_stylemodel(self, conditioning, style_model, clip_vision_output, strength, strength_type):
cond = style_model.get_cond(clip_vision_output).flatten(start_dim=0, end_dim=1).unsqueeze(dim=0)
@ -1521,13 +1521,11 @@ class LatentCrop:
class SetLatentNoiseMask:
@classmethod
def INPUT_TYPES(s):
return {"required": { "samples": ("LATENT",),
"mask": ("MASK",),
}}
return {"required": { "samples": ("LATENT",), "mask": ("MASK",), }}
RETURN_TYPES = ("LATENT",)
FUNCTION = "set_mask"
CATEGORY = "model/latent/inpaint"
CATEGORY = "model/latent"
def set_mask(self, samples, mask):
s = samples.copy()
@ -2053,7 +2051,7 @@ NODE_CLASS_MAPPINGS = {
"ImageBatch": ImageBatch,
"ImagePadForOutpaint": ImagePadForOutpaint,
"EmptyImage": EmptyImage,
"ConditioningAverage": ConditioningAverage ,
"ConditioningAverage": ConditioningAverage,
"ConditioningCombine": ConditioningCombine,
"ConditioningConcat": ConditioningConcat,
"ConditioningSetArea": ConditioningSetArea,
@ -2109,6 +2107,7 @@ NODE_DISPLAY_NAME_MAPPINGS = {
"LoraLoader": "Load LoRA (Model and CLIP)",
"LoraLoaderModelOnly": "Load LoRA",
"CLIPLoader": "Load CLIP",
"DualCLIPLoader": "Load CLIP (Dual)",
"ControlNetLoader": "Load ControlNet Model",
"DiffControlNetLoader": "Load ControlNet Model (diff)",
"StyleModelLoader": "Load Style Model",
@ -2116,6 +2115,7 @@ NODE_DISPLAY_NAME_MAPPINGS = {
"UNETLoader": "Load Diffusion Model",
"unCLIPCheckpointLoader": "Load unCLIP Checkpoint",
"GLIGENLoader": "Load GLIGEN Model",
"DiffusersLoader": "Load Diffusers Model (DEPRECATED)",
# Conditioning
"CLIPVisionEncode": "CLIP Vision Encode",
"StyleModelApply": "Apply Style Model",
@ -2123,12 +2123,16 @@ NODE_DISPLAY_NAME_MAPPINGS = {
"CLIPSetLastLayer": "CLIP Set Last Layer",
"ConditioningCombine": "Conditioning (Combine)",
"ConditioningAverage ": "Conditioning (Average)",
"ConditioningAverage": "Conditioning (Average)",
"ConditioningConcat": "Conditioning (Concat)",
"ConditioningSetArea": "Conditioning (Set Area)",
"ConditioningSetAreaPercentage": "Conditioning (Set Area with Percentage)",
"ConditioningSetAreaStrength": "Conditioning (Set Area Strength)",
"ConditioningSetMask": "Conditioning (Set Mask)",
"ControlNetApply": "Apply ControlNet (DEPRECATED)",
"ControlNetApplyAdvanced": "Apply ControlNet",
"GLIGENTextBoxApply": "Apply GLIGEN Text Box",
"ConditioningZeroOut": "Conditioning Zero Out",
# Latent
"VAEEncodeForInpaint": "VAE Encode (for Inpainting)",
"SetLatentNoiseMask": "Set Latent Noise Mask",
@ -2142,7 +2146,7 @@ NODE_DISPLAY_NAME_MAPPINGS = {
"LatentUpscaleBy": "Upscale Latent By",
"LatentComposite": "Latent Composite",
"LatentBlend": "Latent Blend",
"LatentFromBatch" : "Latent From Batch",
"LatentFromBatch" : "Get Latent From Batch",
"RepeatLatentBatch": "Repeat Latent Batch",
# Image
"EmptyImage": "Empty Image",