Merge 95609acef4 into abe2ec26a6

* feat: Add search_aliases field to node schema

Adds `search_aliases` field to improve node discoverability. Users can define alternative search terms for nodes (e.g., "text concat" → StringConcatenate).

Changes:
- Add `search_aliases: list[str]` to V3 Schema
- Add `SEARCH_ALIASES` support for V1 nodes
- Include field in `/object_info` response
- Add aliases to high-priority core nodes

V1 usage:
```python
class MyNode:
    SEARCH_ALIASES = ["alt name", "synonym"]
```

V3 usage:
```python
io.Schema(
    node_id="MyNode",
    search_aliases=["alt name", "synonym"],
    ...
)
```

## Related PRs
- Frontend: Comfy-Org/ComfyUI_frontend#XXXX (draft - merge after this)
- Docs: Comfy-Org/docs#XXXX (draft - merge after stable)

* Propagate search_aliases through V3 Schema.get_v1_info to NodeInfoV1

.github/ISSUE_TEMPLATE/config.yml

@@ -4,7 +4,7 @@ contact_links:
     url: https://github.com/Comfy-Org/ComfyUI_frontend/issues
     about: Issues related to the ComfyUI frontend (display issues, user interaction bugs), please go to the frontend repo to file the issue
   - name: ComfyUI Matrix Space
-    url: https://app.element.io/#/room/%23comfyui_space%3Amatrix.org
+    url: https://matrix.to/#/#comfyui_space:matrix.org
     about: The ComfyUI Matrix Space is available for support and general discussion related to ComfyUI (Matrix is like Discord but open source).
   - name: Comfy Org Discord
     url: https://discord.gg/comfyorg

.github/ISSUE_TEMPLATE/feature-request.yml

@@ -11,7 +11,7 @@ body:
                 **2:** You have looked to make sure there is not already a feature that does what you need, and there is not already a Feature Request listed for the same idea.
                 **3:** This is something that makes sense to add to ComfyUI Core, and wouldn't make more sense as a custom node.
 
-                If unsure, ask on the [ComfyUI Matrix Space](https://app.element.io/#/room/%23comfyui_space%3Amatrix.org) or the [Comfy Org Discord](https://discord.gg/comfyorg) first.
+                If unsure, ask on the [ComfyUI Matrix Space](https://matrix.to/#/#comfyui_space:matrix.org) or the [Comfy Org Discord](https://discord.gg/comfyorg) first.
     - type: textarea
       attributes:
             label: Feature Idea

.github/ISSUE_TEMPLATE/user-support.yml

@@ -10,7 +10,7 @@ body:
             **1:** You are running the latest version of ComfyUI.
             **2:** You have made an effort to find public answers to your question before asking here. In other words, you googled it first, and scrolled through recent help topics.
 
-                If unsure, ask on the [ComfyUI Matrix Space](https://app.element.io/#/room/%23comfyui_space%3Amatrix.org) or the [Comfy Org Discord](https://discord.gg/comfyorg) first.
+                If unsure, ask on the [ComfyUI Matrix Space](https://matrix.to/#/#comfyui_space:matrix.org) or the [Comfy Org Discord](https://discord.gg/comfyorg) first.
     - type: checkboxes
       id: custom-nodes-test
       attributes:

CONTRIBUTING.md

@@ -6,7 +6,7 @@ There are several ways in which you can contribute, beyond writing code. The goa
 
 ## Asking Questions
 
-Have a question? Instead of opening an issue, please ask on [Discord](https://comfy.org/discord) or [Matrix](https://app.element.io/#/room/%23comfyui_space%3Amatrix.org) channels. Our team and the community will help you.
+Have a question? Instead of opening an issue, please ask on [Discord](https://comfy.org/discord) or [Matrix](https://matrix.to/#/#comfyui_space:matrix.org) channels. Our team and the community will help you.
 
 ## Providing Feedback
 

README.md

@@ -15,7 +15,7 @@
 [![][github-downloads-latest-shield]][github-downloads-link]
 
 [matrix-shield]: https://img.shields.io/badge/Matrix-000000?style=flat&logo=matrix&logoColor=white
-[matrix-url]: https://app.element.io/#/room/%23comfyui_space%3Amatrix.org
+[matrix-url]: https://matrix.to/#/#comfyui_space:matrix.org
 [website-shield]: https://img.shields.io/badge/ComfyOrg-4285F4?style=flat
 [website-url]: https://www.comfy.org/
 <!-- Workaround to display total user from https://github.com/badges/shields/issues/4500#issuecomment-2060079995 -->
@@ -408,7 +408,7 @@ Use `--tls-keyfile key.pem --tls-certfile cert.pem` to enable TLS/SSL, the app w
 
 [Discord](https://comfy.org/discord): Try the #help or #feedback channels.
 
-[Matrix space: #comfyui_space:matrix.org](https://app.element.io/#/room/%23comfyui_space%3Amatrix.org) (it's like discord but open source).
+[Matrix space: #comfyui_space:matrix.org](https://matrix.to/#/#comfyui_space:matrix.org) (it's like discord but open source).
 
 See also: [https://www.comfy.org/](https://www.comfy.org/)
 

comfy/ldm/anima/model.py

@@ -0,0 +1,202 @@
+from comfy.ldm.cosmos.predict2 import MiniTrainDIT
+import torch
+from torch import nn
+import torch.nn.functional as F
+
+
+def rotate_half(x):
+    x1 = x[..., : x.shape[-1] // 2]
+    x2 = x[..., x.shape[-1] // 2 :]
+    return torch.cat((-x2, x1), dim=-1)
+
+
+def apply_rotary_pos_emb(x, cos, sin, unsqueeze_dim=1):
+    cos = cos.unsqueeze(unsqueeze_dim)
+    sin = sin.unsqueeze(unsqueeze_dim)
+    x_embed = (x * cos) + (rotate_half(x) * sin)
+    return x_embed
+
+
+class RotaryEmbedding(nn.Module):
+    def __init__(self, head_dim):
+        super().__init__()
+        self.rope_theta = 10000
+        inv_freq = 1.0 / (self.rope_theta ** (torch.arange(0, head_dim, 2, dtype=torch.int64).to(dtype=torch.float) / head_dim))
+        self.register_buffer("inv_freq", inv_freq, persistent=False)
+
+    @torch.no_grad()
+    def forward(self, x, position_ids):
+        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1).to(x.device)
+        position_ids_expanded = position_ids[:, None, :].float()
+
+        device_type = x.device.type if isinstance(x.device.type, str) and x.device.type != "mps" else "cpu"
+        with torch.autocast(device_type=device_type, enabled=False):  # Force float32
+            freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).transpose(1, 2)
+            emb = torch.cat((freqs, freqs), dim=-1)
+            cos = emb.cos()
+            sin = emb.sin()
+
+        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)
+
+
+class Attention(nn.Module):
+    def __init__(self, query_dim, context_dim, n_heads, head_dim, device=None, dtype=None, operations=None):
+        super().__init__()
+
+        inner_dim = head_dim * n_heads
+        self.n_heads = n_heads
+        self.head_dim = head_dim
+        self.query_dim = query_dim
+        self.context_dim = context_dim
+
+        self.q_proj = operations.Linear(query_dim, inner_dim, bias=False, device=device, dtype=dtype)
+        self.q_norm = operations.RMSNorm(self.head_dim, eps=1e-6, device=device, dtype=dtype)
+
+        self.k_proj = operations.Linear(context_dim, inner_dim, bias=False, device=device, dtype=dtype)
+        self.k_norm = operations.RMSNorm(self.head_dim, eps=1e-6, device=device, dtype=dtype)
+
+        self.v_proj = operations.Linear(context_dim, inner_dim, bias=False, device=device, dtype=dtype)
+
+        self.o_proj = operations.Linear(inner_dim, query_dim, bias=False, device=device, dtype=dtype)
+
+    def forward(self, x, mask=None, context=None, position_embeddings=None, position_embeddings_context=None):
+        context = x if context is None else context
+        input_shape = x.shape[:-1]
+        q_shape = (*input_shape, self.n_heads, self.head_dim)
+        context_shape = context.shape[:-1]
+        kv_shape = (*context_shape, self.n_heads, self.head_dim)
+
+        query_states = self.q_norm(self.q_proj(x).view(q_shape)).transpose(1, 2)
+        key_states = self.k_norm(self.k_proj(context).view(kv_shape)).transpose(1, 2)
+        value_states = self.v_proj(context).view(kv_shape).transpose(1, 2)
+
+        if position_embeddings is not None:
+            assert position_embeddings_context is not None
+            cos, sin = position_embeddings
+            query_states = apply_rotary_pos_emb(query_states, cos, sin)
+            cos, sin = position_embeddings_context
+            key_states = apply_rotary_pos_emb(key_states, cos, sin)
+
+        attn_output = F.scaled_dot_product_attention(query_states, key_states, value_states, attn_mask=mask)
+
+        attn_output = attn_output.transpose(1, 2).reshape(*input_shape, -1).contiguous()
+        attn_output = self.o_proj(attn_output)
+        return attn_output
+
+    def init_weights(self):
+        torch.nn.init.zeros_(self.o_proj.weight)
+
+
+class TransformerBlock(nn.Module):
+    def __init__(self, source_dim, model_dim, num_heads=16, mlp_ratio=4.0, use_self_attn=False, layer_norm=False, device=None, dtype=None, operations=None):
+        super().__init__()
+        self.use_self_attn = use_self_attn
+
+        if self.use_self_attn:
+            self.norm_self_attn = operations.LayerNorm(model_dim, device=device, dtype=dtype) if layer_norm else operations.RMSNorm(model_dim, eps=1e-6, device=device, dtype=dtype)
+            self.self_attn = Attention(
+                query_dim=model_dim,
+                context_dim=model_dim,
+                n_heads=num_heads,
+                head_dim=model_dim//num_heads,
+                device=device,
+                dtype=dtype,
+                operations=operations,
+            )
+
+        self.norm_cross_attn = operations.LayerNorm(model_dim, device=device, dtype=dtype) if layer_norm else operations.RMSNorm(model_dim, eps=1e-6, device=device, dtype=dtype)
+        self.cross_attn = Attention(
+            query_dim=model_dim,
+            context_dim=source_dim,
+            n_heads=num_heads,
+            head_dim=model_dim//num_heads,
+            device=device,
+            dtype=dtype,
+            operations=operations,
+        )
+
+        self.norm_mlp = operations.LayerNorm(model_dim, device=device, dtype=dtype) if layer_norm else operations.RMSNorm(model_dim, eps=1e-6, device=device, dtype=dtype)
+        self.mlp = nn.Sequential(
+            operations.Linear(model_dim, int(model_dim * mlp_ratio), device=device, dtype=dtype),
+            nn.GELU(),
+            operations.Linear(int(model_dim * mlp_ratio), model_dim, device=device, dtype=dtype)
+        )
+
+    def forward(self, x, context, target_attention_mask=None, source_attention_mask=None, position_embeddings=None, position_embeddings_context=None):
+        if self.use_self_attn:
+            normed = self.norm_self_attn(x)
+            attn_out = self.self_attn(normed, mask=target_attention_mask, position_embeddings=position_embeddings, position_embeddings_context=position_embeddings)
+            x = x + attn_out
+
+        normed = self.norm_cross_attn(x)
+        attn_out = self.cross_attn(normed, mask=source_attention_mask, context=context, position_embeddings=position_embeddings, position_embeddings_context=position_embeddings_context)
+        x = x + attn_out
+
+        x = x + self.mlp(self.norm_mlp(x))
+        return x
+
+    def init_weights(self):
+        torch.nn.init.zeros_(self.mlp[2].weight)
+        self.cross_attn.init_weights()
+
+
+class LLMAdapter(nn.Module):
+    def __init__(
+            self,
+            source_dim=1024,
+            target_dim=1024,
+            model_dim=1024,
+            num_layers=6,
+            num_heads=16,
+            use_self_attn=True,
+            layer_norm=False,
+            device=None,
+            dtype=None,
+            operations=None,
+        ):
+        super().__init__()
+
+        self.embed = operations.Embedding(32128, target_dim, device=device, dtype=dtype)
+        if model_dim != target_dim:
+            self.in_proj = operations.Linear(target_dim, model_dim, device=device, dtype=dtype)
+        else:
+            self.in_proj = nn.Identity()
+        self.rotary_emb = RotaryEmbedding(model_dim//num_heads)
+        self.blocks = nn.ModuleList([
+            TransformerBlock(source_dim, model_dim, num_heads=num_heads, use_self_attn=use_self_attn, layer_norm=layer_norm, device=device, dtype=dtype, operations=operations) for _ in range(num_layers)
+        ])
+        self.out_proj = operations.Linear(model_dim, target_dim, device=device, dtype=dtype)
+        self.norm = operations.RMSNorm(target_dim, eps=1e-6, device=device, dtype=dtype)
+
+    def forward(self, source_hidden_states, target_input_ids, target_attention_mask=None, source_attention_mask=None):
+        if target_attention_mask is not None:
+            target_attention_mask = target_attention_mask.to(torch.bool)
+            if target_attention_mask.ndim == 2:
+                target_attention_mask = target_attention_mask.unsqueeze(1).unsqueeze(1)
+
+        if source_attention_mask is not None:
+            source_attention_mask = source_attention_mask.to(torch.bool)
+            if source_attention_mask.ndim == 2:
+                source_attention_mask = source_attention_mask.unsqueeze(1).unsqueeze(1)
+
+        x = self.in_proj(self.embed(target_input_ids))
+        context = source_hidden_states
+        position_ids = torch.arange(x.shape[1], device=x.device).unsqueeze(0)
+        position_ids_context = torch.arange(context.shape[1], device=x.device).unsqueeze(0)
+        position_embeddings = self.rotary_emb(x, position_ids)
+        position_embeddings_context = self.rotary_emb(x, position_ids_context)
+        for block in self.blocks:
+            x = block(x, context, target_attention_mask=target_attention_mask, source_attention_mask=source_attention_mask, position_embeddings=position_embeddings, position_embeddings_context=position_embeddings_context)
+        return self.norm(self.out_proj(x))
+
+
+class Anima(MiniTrainDIT):
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.llm_adapter = LLMAdapter(device=kwargs.get("device"), dtype=kwargs.get("dtype"), operations=kwargs.get("operations"))
+
+    def preprocess_text_embeds(self, text_embeds, text_ids):
+        if text_ids is not None:
+            return self.llm_adapter(text_embeds, text_ids)
+        else:
+            return text_embeds

comfy/model_base.py

@@ -49,6 +49,7 @@ import comfy.ldm.ace.model
 import comfy.ldm.omnigen.omnigen2
 import comfy.ldm.qwen_image.model
 import comfy.ldm.kandinsky5.model
+import comfy.ldm.anima.model
 
 import comfy.model_management
 import comfy.patcher_extension
@@ -1147,6 +1148,27 @@ class CosmosPredict2(BaseModel):
         sigma = (sigma / (sigma + 1))
         return latent_image / (1.0 - sigma)
 
+class Anima(BaseModel):
+    def __init__(self, model_config, model_type=ModelType.FLOW, device=None):
+        super().__init__(model_config, model_type, device=device, unet_model=comfy.ldm.anima.model.Anima)
+
+    def extra_conds(self, **kwargs):
+        out = super().extra_conds(**kwargs)
+        cross_attn = kwargs.get("cross_attn", None)
+        t5xxl_ids = kwargs.get("t5xxl_ids", None)
+        t5xxl_weights = kwargs.get("t5xxl_weights", None)
+        device = kwargs["device"]
+        if cross_attn is not None:
+            if t5xxl_ids is not None:
+                cross_attn = self.diffusion_model.preprocess_text_embeds(cross_attn.to(device=device, dtype=self.get_dtype()), t5xxl_ids.unsqueeze(0).to(device=device))
+                if t5xxl_weights is not None:
+                    cross_attn *= t5xxl_weights.unsqueeze(0).unsqueeze(-1).to(cross_attn)
+
+                if cross_attn.shape[1] < 512:
+                    cross_attn = torch.nn.functional.pad(cross_attn, (0, 0, 0, 512 - cross_attn.shape[1]))
+            out['c_crossattn'] = comfy.conds.CONDRegular(cross_attn)
+        return out
+
 class Lumina2(BaseModel):
     def __init__(self, model_config, model_type=ModelType.FLOW, device=None):
         super().__init__(model_config, model_type, device=device, unet_model=comfy.ldm.lumina.model.NextDiT)

comfy/model_detection.py

@@ -550,6 +550,8 @@ def detect_unet_config(state_dict, key_prefix, metadata=None):
     if '{}blocks.0.mlp.layer1.weight'.format(key_prefix) in state_dict_keys:  # Cosmos predict2
         dit_config = {}
         dit_config["image_model"] = "cosmos_predict2"
+        if "{}llm_adapter.blocks.0.cross_attn.q_proj.weight".format(key_prefix) in state_dict_keys:
+            dit_config["image_model"] = "anima"
         dit_config["max_img_h"] = 240
         dit_config["max_img_w"] = 240
         dit_config["max_frames"] = 128

comfy/sd.py

@@ -57,6 +57,7 @@ import comfy.text_encoders.ovis
 import comfy.text_encoders.kandinsky5
 import comfy.text_encoders.jina_clip_2
 import comfy.text_encoders.newbie
+import comfy.text_encoders.anima
 
 import comfy.model_patcher
 import comfy.lora
@@ -1048,6 +1049,7 @@ class TEModel(Enum):
     GEMMA_3_12B = 18
     JINA_CLIP_2 = 19
     QWEN3_8B = 20
+    QWEN3_06B = 21
 
 
 def detect_te_model(sd):
@@ -1093,6 +1095,8 @@ def detect_te_model(sd):
                 return TEModel.QWEN3_2B
             elif weight.shape[0] == 4096:
                 return TEModel.QWEN3_8B
+            elif weight.shape[0] == 1024:
+                return TEModel.QWEN3_06B
         if weight.shape[0] == 5120:
             if "model.layers.39.post_attention_layernorm.weight" in sd:
                 return TEModel.MISTRAL3_24B
@@ -1233,6 +1237,9 @@ def load_text_encoder_state_dicts(state_dicts=[], embedding_directory=None, clip
         elif te_model == TEModel.JINA_CLIP_2:
             clip_target.clip = comfy.text_encoders.jina_clip_2.JinaClip2TextModelWrapper
             clip_target.tokenizer = comfy.text_encoders.jina_clip_2.JinaClip2TokenizerWrapper
+        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
         else:
             # clip_l
             if clip_type == CLIPType.SD3:

comfy/supported_models.py

@@ -23,6 +23,7 @@ import comfy.text_encoders.qwen_image
 import comfy.text_encoders.hunyuan_image
 import comfy.text_encoders.kandinsky5
 import comfy.text_encoders.z_image
+import comfy.text_encoders.anima
 
 from . import supported_models_base
 from . import latent_formats
@@ -992,6 +993,36 @@ class CosmosT2IPredict2(supported_models_base.BASE):
         t5_detect = comfy.text_encoders.sd3_clip.t5_xxl_detect(state_dict, "{}t5xxl.transformer.".format(pref))
         return supported_models_base.ClipTarget(comfy.text_encoders.cosmos.CosmosT5Tokenizer, comfy.text_encoders.cosmos.te(**t5_detect))
 
+class Anima(supported_models_base.BASE):
+    unet_config = {
+        "image_model": "anima",
+    }
+
+    sampling_settings = {
+        "multiplier": 1.0,
+        "shift": 3.0,
+    }
+
+    unet_extra_config = {}
+    latent_format = latent_formats.Wan21
+
+    memory_usage_factor = 1.0
+
+    supported_inference_dtypes = [torch.bfloat16, torch.float32]
+
+    def __init__(self, unet_config):
+        super().__init__(unet_config)
+        self.memory_usage_factor = (unet_config.get("model_channels", 2048) / 2048) * 0.95
+
+    def get_model(self, state_dict, prefix="", device=None):
+        out = model_base.Anima(self, device=device)
+        return out
+
+    def clip_target(self, state_dict={}):
+        pref = self.text_encoder_key_prefix[0]
+        detect = comfy.text_encoders.hunyuan_video.llama_detect(state_dict, "{}qwen3_06b.transformer.".format(pref))
+        return supported_models_base.ClipTarget(comfy.text_encoders.anima.AnimaTokenizer, comfy.text_encoders.anima.te(**detect))
+
 class CosmosI2VPredict2(CosmosT2IPredict2):
     unet_config = {
         "image_model": "cosmos_predict2",
@@ -1551,6 +1582,6 @@ class Kandinsky5Image(Kandinsky5):
         return supported_models_base.ClipTarget(comfy.text_encoders.kandinsky5.Kandinsky5TokenizerImage, comfy.text_encoders.kandinsky5.te(**hunyuan_detect))
 
 
-models = [LotusD, Stable_Zero123, SD15_instructpix2pix, SD15, SD20, SD21UnclipL, SD21UnclipH, SDXL_instructpix2pix, SDXLRefiner, SDXL, SSD1B, KOALA_700M, KOALA_1B, Segmind_Vega, SD_X4Upscaler, Stable_Cascade_C, Stable_Cascade_B, SV3D_u, SV3D_p, SD3, StableAudio, AuraFlow, PixArtAlpha, PixArtSigma, HunyuanDiT, HunyuanDiT1, FluxInpaint, Flux, FluxSchnell, GenmoMochi, LTXV, LTXAV, HunyuanVideo15_SR_Distilled, HunyuanVideo15, HunyuanImage21Refiner, HunyuanImage21, HunyuanVideoSkyreelsI2V, HunyuanVideoI2V, HunyuanVideo, CosmosT2V, CosmosI2V, CosmosT2IPredict2, CosmosI2VPredict2, ZImage, Lumina2, WAN22_T2V, WAN21_T2V, WAN21_I2V, WAN21_FunControl2V, WAN21_Vace, WAN21_Camera, WAN22_Camera, WAN22_S2V, WAN21_HuMo, WAN22_Animate, Hunyuan3Dv2mini, Hunyuan3Dv2, Hunyuan3Dv2_1, HiDream, Chroma, ChromaRadiance, ACEStep, Omnigen2, QwenImage, Flux2, Kandinsky5Image, Kandinsky5]
+models = [LotusD, Stable_Zero123, SD15_instructpix2pix, SD15, SD20, SD21UnclipL, SD21UnclipH, SDXL_instructpix2pix, SDXLRefiner, SDXL, SSD1B, KOALA_700M, KOALA_1B, Segmind_Vega, SD_X4Upscaler, Stable_Cascade_C, Stable_Cascade_B, SV3D_u, SV3D_p, SD3, StableAudio, AuraFlow, PixArtAlpha, PixArtSigma, HunyuanDiT, HunyuanDiT1, FluxInpaint, Flux, FluxSchnell, GenmoMochi, LTXV, LTXAV, HunyuanVideo15_SR_Distilled, HunyuanVideo15, HunyuanImage21Refiner, HunyuanImage21, HunyuanVideoSkyreelsI2V, HunyuanVideoI2V, HunyuanVideo, CosmosT2V, CosmosI2V, CosmosT2IPredict2, CosmosI2VPredict2, ZImage, Lumina2, WAN22_T2V, WAN21_T2V, WAN21_I2V, WAN21_FunControl2V, WAN21_Vace, WAN21_Camera, WAN22_Camera, WAN22_S2V, WAN21_HuMo, WAN22_Animate, Hunyuan3Dv2mini, Hunyuan3Dv2, Hunyuan3Dv2_1, HiDream, Chroma, ChromaRadiance, ACEStep, Omnigen2, QwenImage, Flux2, Kandinsky5Image, Kandinsky5, Anima]
 
 models += [SVD_img2vid]

comfy/text_encoders/anima.py

@@ -0,0 +1,61 @@
+from transformers import Qwen2Tokenizer, T5TokenizerFast
+import comfy.text_encoders.llama
+from comfy import sd1_clip
+import os
+import torch
+
+
+class Qwen3Tokenizer(sd1_clip.SDTokenizer):
+    def __init__(self, embedding_directory=None, tokenizer_data={}):
+        tokenizer_path = os.path.join(os.path.dirname(os.path.realpath(__file__)), "qwen25_tokenizer")
+        super().__init__(tokenizer_path, pad_with_end=False, embedding_size=1024, embedding_key='qwen3_06b', 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 T5XXLTokenizer(sd1_clip.SDTokenizer):
+    def __init__(self, embedding_directory=None, tokenizer_data={}):
+        tokenizer_path = os.path.join(os.path.dirname(os.path.realpath(__file__)), "t5_tokenizer")
+        super().__init__(tokenizer_path, embedding_directory=embedding_directory, pad_with_end=False, embedding_size=4096, embedding_key='t5xxl', tokenizer_class=T5TokenizerFast, has_start_token=False, pad_to_max_length=False, max_length=99999999, min_length=1, tokenizer_data=tokenizer_data)
+
+class AnimaTokenizer:
+    def __init__(self, embedding_directory=None, tokenizer_data={}):
+        self.qwen3_06b = Qwen3Tokenizer(embedding_directory=embedding_directory, tokenizer_data=tokenizer_data)
+        self.t5xxl = T5XXLTokenizer(embedding_directory=embedding_directory, tokenizer_data=tokenizer_data)
+
+    def tokenize_with_weights(self, text:str, return_word_ids=False, **kwargs):
+        out = {}
+        qwen_ids = self.qwen3_06b.tokenize_with_weights(text, return_word_ids, **kwargs)
+        out["qwen3_06b"] = [[(token, 1.0) for token, _ in inner_list] for inner_list in qwen_ids]  # Set weights to 1.0
+        out["t5xxl"] = self.t5xxl.tokenize_with_weights(text, return_word_ids, **kwargs)
+        return out
+
+    def untokenize(self, token_weight_pair):
+        return self.t5xxl.untokenize(token_weight_pair)
+
+    def state_dict(self):
+        return {}
+
+
+class Qwen3_06BModel(sd1_clip.SDClipModel):
+    def __init__(self, device="cpu", layer="last", layer_idx=None, dtype=None, attention_mask=True, model_options={}):
+        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=comfy.text_encoders.llama.Qwen3_06B, enable_attention_masks=attention_mask, return_attention_masks=attention_mask, model_options=model_options)
+
+
+class AnimaTEModel(sd1_clip.SD1ClipModel):
+    def __init__(self, device="cpu", dtype=None, model_options={}):
+        super().__init__(device=device, dtype=dtype, name="qwen3_06b", clip_model=Qwen3_06BModel, model_options=model_options)
+
+    def encode_token_weights(self, token_weight_pairs):
+        out = super().encode_token_weights(token_weight_pairs)
+        out[2]["t5xxl_ids"] = torch.tensor(list(map(lambda a: a[0], token_weight_pairs["t5xxl"][0])), dtype=torch.int)
+        out[2]["t5xxl_weights"] = torch.tensor(list(map(lambda a: a[1], token_weight_pairs["t5xxl"][0])))
+        return out
+
+def te(dtype_llama=None, llama_quantization_metadata=None):
+    class AnimaTEModel_(AnimaTEModel):
+        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 AnimaTEModel_

comfy/text_encoders/llama.py

@@ -77,6 +77,28 @@ class Qwen25_3BConfig:
     rope_scale = None
     final_norm: bool = True
 
+@dataclass
+class Qwen3_06BConfig:
+    vocab_size: int = 151936
+    hidden_size: int = 1024
+    intermediate_size: int = 3072
+    num_hidden_layers: int = 28
+    num_attention_heads: int = 16
+    num_key_value_heads: int = 8
+    max_position_embeddings: int = 32768
+    rms_norm_eps: float = 1e-6
+    rope_theta: float = 1000000.0
+    transformer_type: str = "llama"
+    head_dim = 128
+    rms_norm_add = False
+    mlp_activation = "silu"
+    qkv_bias = False
+    rope_dims = None
+    q_norm = "gemma3"
+    k_norm = "gemma3"
+    rope_scale = None
+    final_norm: bool = True
+
 @dataclass
 class Qwen3_4BConfig:
     vocab_size: int = 151936
@@ -641,6 +663,15 @@ class Qwen25_3B(BaseLlama, torch.nn.Module):
         self.model = Llama2_(config, device=device, dtype=dtype, ops=operations)
         self.dtype = dtype
 
+class Qwen3_06B(BaseLlama, torch.nn.Module):
+    def __init__(self, config_dict, dtype, device, operations):
+        super().__init__()
+        config = Qwen3_06BConfig(**config_dict)
+        self.num_layers = config.num_hidden_layers
+
+        self.model = Llama2_(config, device=device, dtype=dtype, ops=operations)
+        self.dtype = dtype
+
 class Qwen3_4B(BaseLlama, torch.nn.Module):
     def __init__(self, config_dict, dtype, device, operations):
         super().__init__()

comfy/utils.py

@@ -611,6 +611,14 @@ def flux_to_diffusers(mmdit_config, output_prefix=""):
                         "ff_context.net.0.proj.bias": "txt_mlp.0.bias",
                         "ff_context.net.2.weight": "txt_mlp.2.weight",
                         "ff_context.net.2.bias": "txt_mlp.2.bias",
+                        "ff.linear_in.weight": "img_mlp.0.weight",  # LyCoris LoKr
+                        "ff.linear_in.bias": "img_mlp.0.bias",
+                        "ff.linear_out.weight": "img_mlp.2.weight",
+                        "ff.linear_out.bias": "img_mlp.2.bias",
+                        "ff_context.linear_in.weight": "txt_mlp.0.weight",
+                        "ff_context.linear_in.bias": "txt_mlp.0.bias",
+                        "ff_context.linear_out.weight": "txt_mlp.2.weight",
+                        "ff_context.linear_out.bias": "txt_mlp.2.bias",
                         "attn.norm_q.weight": "img_attn.norm.query_norm.scale",
                         "attn.norm_k.weight": "img_attn.norm.key_norm.scale",
                         "attn.norm_added_q.weight": "txt_attn.norm.query_norm.scale",

comfy_api/latest/_io.py

@@ -1249,6 +1249,7 @@ class NodeInfoV1:
     experimental: bool=None
     api_node: bool=None
     price_badge: dict | None = None
+    search_aliases: list[str]=None
 
 @dataclass
 class NodeInfoV3:
@@ -1346,6 +1347,8 @@ class Schema:
     hidden: list[Hidden] = field(default_factory=list)
     description: str=""
     """Node description, shown as a tooltip when hovering over the node."""
+    search_aliases: list[str] = field(default_factory=list)
+    """Alternative names for search. Useful for synonyms, abbreviations, or old names after renaming."""
     is_input_list: bool = False
     """A flag indicating if this node implements the additional code necessary to deal with OUTPUT_IS_LIST nodes.
 
@@ -1483,6 +1486,7 @@ class Schema:
             api_node=self.is_api_node,
             python_module=getattr(cls, "RELATIVE_PYTHON_MODULE", "nodes"),
             price_badge=self.price_badge.as_dict(self.inputs) if self.price_badge is not None else None,
+            search_aliases=self.search_aliases if self.search_aliases else None,
         )
         return info
 

comfy_api_nodes/nodes_vidu.py

@@ -703,7 +703,7 @@ class Vidu2ReferenceVideoNode(IO.ComfyNode):
                     "subjects",
                     template=IO.Autogrow.TemplateNames(
                         IO.Image.Input("reference_images"),
-                        names=["subject1", "subject2", "subject3"],
+                        names=["subject1", "subject2", "subject3", "subject4", "subject5", "subject6", "subject7"],
                         min=1,
                     ),
                     tooltip="For each subject, provide up to 3 reference images (7 images total across all subjects). "
@@ -738,7 +738,7 @@ class Vidu2ReferenceVideoNode(IO.ComfyNode):
                     control_after_generate=True,
                 ),
                 IO.Combo.Input("aspect_ratio", options=["16:9", "9:16", "4:3", "3:4", "1:1"]),
-                IO.Combo.Input("resolution", options=["720p"]),
+                IO.Combo.Input("resolution", options=["720p", "1080p"]),
                 IO.Combo.Input(
                     "movement_amplitude",
                     options=["auto", "small", "medium", "large"],

comfy_extras/nodes_post_processing.py

@@ -550,6 +550,7 @@ class BatchImagesNode(io.ComfyNode):
             node_id="BatchImagesNode",
             display_name="Batch Images",
             category="image",
+            search_aliases=["batch", "image batch", "batch images", "combine images", "merge images", "stack images"],
             inputs=[
                 io.Autogrow.Input("images", template=autogrow_template)
             ],

comfy_extras/nodes_preview_any.py

@@ -16,6 +16,7 @@ class PreviewAny():
     OUTPUT_NODE = True
 
     CATEGORY = "utils"
+    SEARCH_ALIASES = ["preview", "show", "display", "view", "show text", "display text", "preview text", "show output", "inspect", "debug"]
 
     def main(self, source=None):
         value = 'None'

comfy_extras/nodes_string.py

@@ -11,6 +11,7 @@ class StringConcatenate(io.ComfyNode):
             node_id="StringConcatenate",
             display_name="Concatenate",
             category="utils/string",
+            search_aliases=["text concat", "join text", "merge text", "combine strings", "concat", "concatenate", "append text", "combine text", "string"],
             inputs=[
                 io.String.Input("string_a", multiline=True),
                 io.String.Input("string_b", multiline=True),

comfy_extras/nodes_upscale_model.py

@@ -53,6 +53,7 @@ class ImageUpscaleWithModel(io.ComfyNode):
             node_id="ImageUpscaleWithModel",
             display_name="Upscale Image (using Model)",
             category="image/upscaling",
+            search_aliases=["upscale", "upscaler", "upsc", "enlarge image", "super resolution", "hires", "superres", "increase resolution"],
             inputs=[
                 io.UpscaleModel.Input("upscale_model"),
                 io.Image.Input("image"),

nodes.py

@@ -70,6 +70,7 @@ class CLIPTextEncode(ComfyNodeABC):
 
     CATEGORY = "conditioning"
     DESCRIPTION = "Encodes a text prompt using a CLIP model into an embedding that can be used to guide the diffusion model towards generating specific images."
+    SEARCH_ALIASES = ["text", "prompt", "text prompt", "positive prompt", "negative prompt", "encode text", "text encoder", "encode prompt"]
 
     def encode(self, clip, text):
         if clip is None:
@@ -86,6 +87,7 @@ class ConditioningCombine:
     FUNCTION = "combine"
 
     CATEGORY = "conditioning"
+    SEARCH_ALIASES = ["combine", "merge conditioning", "combine prompts", "merge prompts", "mix prompts", "add prompt"]
 
     def combine(self, conditioning_1, conditioning_2):
         return (conditioning_1 + conditioning_2, )
@@ -294,6 +296,7 @@ class VAEDecode:
 
     CATEGORY = "latent"
     DESCRIPTION = "Decodes latent images back into pixel space images."
+    SEARCH_ALIASES = ["decode", "decode latent", "latent to image", "render latent"]
 
     def decode(self, vae, samples):
         latent = samples["samples"]
@@ -346,6 +349,7 @@ class VAEEncode:
     FUNCTION = "encode"
 
     CATEGORY = "latent"
+    SEARCH_ALIASES = ["encode", "encode image", "image to latent"]
 
     def encode(self, vae, pixels):
         t = vae.encode(pixels)
@@ -581,6 +585,7 @@ class CheckpointLoaderSimple:
 
     CATEGORY = "loaders"
     DESCRIPTION = "Loads a diffusion model checkpoint, diffusion models are used to denoise latents."
+    SEARCH_ALIASES = ["load model", "checkpoint", "model loader", "load checkpoint", "ckpt", "model"]
 
     def load_checkpoint(self, ckpt_name):
         ckpt_path = folder_paths.get_full_path_or_raise("checkpoints", ckpt_name)
@@ -667,6 +672,7 @@ class LoraLoader:
 
     CATEGORY = "loaders"
     DESCRIPTION = "LoRAs are used to modify diffusion and CLIP models, altering the way in which latents are denoised such as applying styles. Multiple LoRA nodes can be linked together."
+    SEARCH_ALIASES = ["lora", "load lora", "apply lora", "lora loader", "lora model"]
 
     def load_lora(self, model, clip, lora_name, strength_model, strength_clip):
         if strength_model == 0 and strength_clip == 0:
@@ -814,6 +820,7 @@ class ControlNetLoader:
     FUNCTION = "load_controlnet"
 
     CATEGORY = "loaders"
+    SEARCH_ALIASES = ["controlnet", "control net", "cn", "load controlnet", "controlnet loader"]
 
     def load_controlnet(self, control_net_name):
         controlnet_path = folder_paths.get_full_path_or_raise("controlnet", control_net_name)
@@ -890,6 +897,7 @@ class ControlNetApplyAdvanced:
     FUNCTION = "apply_controlnet"
 
     CATEGORY = "conditioning/controlnet"
+    SEARCH_ALIASES = ["controlnet", "apply controlnet", "use controlnet", "control net"]
 
     def apply_controlnet(self, positive, negative, control_net, image, strength, start_percent, end_percent, vae=None, extra_concat=[]):
         if strength == 0:
@@ -1200,6 +1208,7 @@ class EmptyLatentImage:
 
     CATEGORY = "latent"
     DESCRIPTION = "Create a new batch of empty latent images to be denoised via sampling."
+    SEARCH_ALIASES = ["empty", "empty latent", "new latent", "create latent", "blank latent", "blank"]
 
     def generate(self, width, height, batch_size=1):
         latent = torch.zeros([batch_size, 4, height // 8, width // 8], device=self.device)
@@ -1540,6 +1549,7 @@ class KSampler:
 
     CATEGORY = "sampling"
     DESCRIPTION = "Uses the provided model, positive and negative conditioning to denoise the latent image."
+    SEARCH_ALIASES = ["sampler", "sample", "generate", "denoise", "diffuse", "txt2img", "img2img"]
 
     def sample(self, model, seed, steps, cfg, sampler_name, scheduler, positive, negative, latent_image, denoise=1.0):
         return common_ksampler(model, seed, steps, cfg, sampler_name, scheduler, positive, negative, latent_image, denoise=denoise)
@@ -1604,6 +1614,7 @@ class SaveImage:
 
     CATEGORY = "image"
     DESCRIPTION = "Saves the input images to your ComfyUI output directory."
+    SEARCH_ALIASES = ["save", "save image", "export image", "output image", "write image", "download"]
 
     def save_images(self, images, filename_prefix="ComfyUI", prompt=None, extra_pnginfo=None):
         filename_prefix += self.prefix_append
@@ -1640,6 +1651,8 @@ class PreviewImage(SaveImage):
         self.prefix_append = "_temp_" + ''.join(random.choice("abcdefghijklmnopqrstupvxyz") for x in range(5))
         self.compress_level = 1
 
+    SEARCH_ALIASES = ["preview", "preview image", "show image", "view image", "display image", "image viewer"]
+
     @classmethod
     def INPUT_TYPES(s):
         return {"required":
@@ -1658,6 +1671,7 @@ class LoadImage:
                 }
 
     CATEGORY = "image"
+    SEARCH_ALIASES = ["load image", "open image", "import image", "image input", "upload image", "read image", "image loader"]
 
     RETURN_TYPES = ("IMAGE", "MASK")
     FUNCTION = "load_image"
@@ -1810,6 +1824,7 @@ class ImageScale:
     FUNCTION = "upscale"
 
     CATEGORY = "image/upscaling"
+    SEARCH_ALIASES = ["resize", "resize image", "scale image", "image resize", "zoom", "zoom in", "change size"]
 
     def upscale(self, image, upscale_method, width, height, crop):
         if width == 0 and height == 0:

server.py

@@ -682,6 +682,8 @@ class PromptServer():
 
             if hasattr(obj_class, 'API_NODE'):
                 info['api_node'] = obj_class.API_NODE
+
+            info['search_aliases'] = getattr(obj_class, 'SEARCH_ALIASES', [])
             return info
 
         @routes.get("/object_info")