Merge branch 'comfyanonymous:master' into master

comfy/ldm/lightricks/model.py

@@ -379,6 +379,7 @@ class LTXVModel(torch.nn.Module):
                  positional_embedding_max_pos=[20, 2048, 2048],
                  dtype=None, device=None, operations=None, **kwargs):
         super().__init__()
+        self.generator = None
         self.dtype = dtype
         self.out_channels = in_channels
         self.inner_dim = num_attention_heads * attention_head_dim
@@ -415,7 +416,7 @@ class LTXVModel(torch.nn.Module):
 
         self.patchifier = SymmetricPatchifier(1)
 
-    def forward(self, x, timestep, context, attention_mask, frame_rate=25, guiding_latent=None, transformer_options={}, **kwargs):
+    def forward(self, x, timestep, context, attention_mask, frame_rate=25, guiding_latent=None, guiding_latent_noise_scale=0, transformer_options={}, **kwargs):
         patches_replace = transformer_options.get("patches_replace", {})
 
         indices_grid = self.patchifier.get_grid(
@@ -431,10 +432,22 @@ class LTXVModel(torch.nn.Module):
             ts = torch.ones([x.shape[0], 1, x.shape[2], x.shape[3], x.shape[4]], device=x.device, dtype=x.dtype)
             input_ts = timestep.view([timestep.shape[0]] + [1] * (x.ndim - 1))
             ts *= input_ts
-            ts[:, :, 0] = 0.0
+            ts[:, :, 0] = guiding_latent_noise_scale * (input_ts[:, :, 0] ** 2)
             timestep = self.patchifier.patchify(ts)
             input_x = x.clone()
             x[:, :, 0] = guiding_latent[:, :, 0]
+            if guiding_latent_noise_scale > 0:
+                if self.generator is None:
+                    self.generator = torch.Generator(device=x.device).manual_seed(42)
+                elif self.generator.device != x.device:
+                    self.generator = torch.Generator(device=x.device).set_state(self.generator.get_state())
+
+                noise_shape = [guiding_latent.shape[0], guiding_latent.shape[1], 1, guiding_latent.shape[3], guiding_latent.shape[4]]
+                scale = guiding_latent_noise_scale * (input_ts ** 2)
+                guiding_noise = scale * torch.randn(size=noise_shape, device=x.device, generator=self.generator)
+
+                x[:, :, 0] = guiding_noise[:, :, 0] + x[:, :, 0] *  (1.0 - scale[:, :, 0])
+
 
         orig_shape = list(x.shape)
 

comfy/model_base.py

@@ -804,5 +804,9 @@ class LTXV(BaseModel):
         if guiding_latent is not None:
             out['guiding_latent'] = comfy.conds.CONDRegular(guiding_latent)
 
+        guiding_latent_noise_scale = kwargs.get("guiding_latent_noise_scale", None)
+        if guiding_latent_noise_scale is not None:
+            out["guiding_latent_noise_scale"] = comfy.conds.CONDConstant(guiding_latent_noise_scale)
+
         out['frame_rate'] = comfy.conds.CONDConstant(kwargs.get("frame_rate", 25))
         return out

comfy/model_sampling.py

@@ -243,7 +243,7 @@ class ModelSamplingDiscreteFlow(torch.nn.Module):
             return 1.0
         if percent >= 1.0:
             return 0.0
-        return 1.0 - percent
+        return time_snr_shift(self.shift, 1.0 - percent)
 
 class StableCascadeSampling(ModelSamplingDiscrete):
     def __init__(self, model_config=None):
@@ -336,4 +336,4 @@ class ModelSamplingFlux(torch.nn.Module):
             return 1.0
         if percent >= 1.0:
             return 0.0
-        return 1.0 - percent
+        return flux_time_shift(self.shift, 1.0, 1.0 - percent)

comfy_extras/nodes_lt.py

@@ -32,7 +32,9 @@ class LTXVImgToVideo:
                              "width": ("INT", {"default": 768, "min": 64, "max": nodes.MAX_RESOLUTION, "step": 32}),
                              "height": ("INT", {"default": 512, "min": 64, "max": nodes.MAX_RESOLUTION, "step": 32}),
                              "length": ("INT", {"default": 97, "min": 9, "max": nodes.MAX_RESOLUTION, "step": 8}),
-                             "batch_size": ("INT", {"default": 1, "min": 1, "max": 4096})}}
+                             "batch_size": ("INT", {"default": 1, "min": 1, "max": 4096}),
+                             "image_noise_scale": ("FLOAT", {"default": 0.15, "min": 0, "max": 1.0, "step": 0.01, "tooltip": "Amount of noise to apply on conditioning image latent."})
+                             }}
 
     RETURN_TYPES = ("CONDITIONING", "CONDITIONING", "LATENT")
     RETURN_NAMES = ("positive", "negative", "latent")
@@ -40,12 +42,12 @@ class LTXVImgToVideo:
     CATEGORY = "conditioning/video_models"
     FUNCTION = "generate"
 
-    def generate(self, positive, negative, image, vae, width, height, length, batch_size):
+    def generate(self, positive, negative, image, vae, width, height, length, batch_size, image_noise_scale):
         pixels = comfy.utils.common_upscale(image.movedim(-1, 1), width, height, "bilinear", "center").movedim(1, -1)
         encode_pixels = pixels[:, :, :, :3]
         t = vae.encode(encode_pixels)
-        positive = node_helpers.conditioning_set_values(positive, {"guiding_latent": t})
+        positive = node_helpers.conditioning_set_values(positive, {"guiding_latent": t, "guiding_latent_noise_scale": image_noise_scale})
-        negative = node_helpers.conditioning_set_values(negative, {"guiding_latent": t})
+        negative = node_helpers.conditioning_set_values(negative, {"guiding_latent": t, "guiding_latent_noise_scale": image_noise_scale})
 
         latent = torch.zeros([batch_size, 128, ((length - 1) // 8) + 1, height // 32, width // 32], device=comfy.model_management.intermediate_device())
         latent[:, :, :t.shape[2]] = t
@@ -109,6 +111,7 @@ class ModelSamplingLTXV:
         model_sampling = ModelSamplingAdvanced(model.model.model_config)
         model_sampling.set_parameters(shift=shift)
         m.add_object_patch("model_sampling", model_sampling)
+
         return (m, )