Support for WAN VACE preview model. (#7711)

* Support for WAN VACE preview model.

* Remove print.

comfy/ldm/wan/model.py

@@ -220,6 +220,34 @@ class WanAttentionBlock(nn.Module):
         return x
 
 
+class VaceWanAttentionBlock(WanAttentionBlock):
+    def __init__(
+            self,
+            cross_attn_type,
+            dim,
+            ffn_dim,
+            num_heads,
+            window_size=(-1, -1),
+            qk_norm=True,
+            cross_attn_norm=False,
+            eps=1e-6,
+            block_id=0,
+            operation_settings={}
+    ):
+        super().__init__(cross_attn_type, dim, ffn_dim, num_heads, window_size, qk_norm, cross_attn_norm, eps, operation_settings=operation_settings)
+        self.block_id = block_id
+        if block_id == 0:
+            self.before_proj = operation_settings.get("operations").Linear(self.dim, self.dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype"))
+        self.after_proj = operation_settings.get("operations").Linear(self.dim, self.dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype"))
+
+    def forward(self, c, x, **kwargs):
+        if self.block_id == 0:
+            c = self.before_proj(c) + x
+        c = super().forward(c, **kwargs)
+        c_skip = self.after_proj(c)
+        return c_skip, c
+
+
 class Head(nn.Module):
 
     def __init__(self, dim, out_dim, patch_size, eps=1e-6, operation_settings={}):
@@ -395,6 +423,7 @@ class WanModel(torch.nn.Module):
         clip_fea=None,
         freqs=None,
         transformer_options={},
+        **kwargs,
     ):
         r"""
         Forward pass through the diffusion model
@@ -457,7 +486,7 @@ class WanModel(torch.nn.Module):
         x = self.unpatchify(x, grid_sizes)
         return x
 
-    def forward(self, x, timestep, context, clip_fea=None, transformer_options={},**kwargs):
+    def forward(self, x, timestep, context, clip_fea=None, transformer_options={}, **kwargs):
         bs, c, t, h, w = x.shape
         x = comfy.ldm.common_dit.pad_to_patch_size(x, self.patch_size)
         patch_size = self.patch_size
@@ -471,7 +500,7 @@ class WanModel(torch.nn.Module):
         img_ids = repeat(img_ids, "t h w c -> b (t h w) c", b=bs)
 
         freqs = self.rope_embedder(img_ids).movedim(1, 2)
-        return self.forward_orig(x, timestep, context, clip_fea=clip_fea, freqs=freqs, transformer_options=transformer_options)[:, :, :t, :h, :w]
+        return self.forward_orig(x, timestep, context, clip_fea=clip_fea, freqs=freqs, transformer_options=transformer_options, **kwargs)[:, :, :t, :h, :w]
 
     def unpatchify(self, x, grid_sizes):
         r"""
@@ -496,3 +525,114 @@ class WanModel(torch.nn.Module):
         u = torch.einsum('bfhwpqrc->bcfphqwr', u)
         u = u.reshape(b, c, *[i * j for i, j in zip(grid_sizes, self.patch_size)])
         return u
+
+
+class VaceWanModel(WanModel):
+    r"""
+    Wan diffusion backbone supporting both text-to-video and image-to-video.
+    """
+
+    def __init__(self,
+                 model_type='vace',
+                 patch_size=(1, 2, 2),
+                 text_len=512,
+                 in_dim=16,
+                 dim=2048,
+                 ffn_dim=8192,
+                 freq_dim=256,
+                 text_dim=4096,
+                 out_dim=16,
+                 num_heads=16,
+                 num_layers=32,
+                 window_size=(-1, -1),
+                 qk_norm=True,
+                 cross_attn_norm=True,
+                 eps=1e-6,
+                 flf_pos_embed_token_number=None,
+                 image_model=None,
+                 vace_layers=None,
+                 vace_in_dim=None,
+                 device=None,
+                 dtype=None,
+                 operations=None,
+                 ):
+
+        super().__init__(model_type='t2v', patch_size=patch_size, text_len=text_len, in_dim=in_dim, dim=dim, ffn_dim=ffn_dim, freq_dim=freq_dim, text_dim=text_dim, out_dim=out_dim, num_heads=num_heads, num_layers=num_layers, window_size=window_size, qk_norm=qk_norm, cross_attn_norm=cross_attn_norm, eps=eps, flf_pos_embed_token_number=flf_pos_embed_token_number, image_model=image_model, device=device, dtype=dtype, operations=operations)
+        operation_settings = {"operations": operations, "device": device, "dtype": dtype}
+
+        # Vace
+        if vace_layers is not None:
+            self.vace_layers = vace_layers
+            self.vace_in_dim = vace_in_dim
+            # vace blocks
+            self.vace_blocks = nn.ModuleList([
+                VaceWanAttentionBlock('t2v_cross_attn', self.dim, self.ffn_dim, self.num_heads, self.window_size, self.qk_norm, self.cross_attn_norm, self.eps, block_id=i, operation_settings=operation_settings)
+                for i in range(self.vace_layers)
+            ])
+
+            self.vace_layers_mapping = {i: n for n, i in enumerate(range(0, self.num_layers, self.num_layers // self.vace_layers))}
+            # vace patch embeddings
+            self.vace_patch_embedding = operations.Conv3d(
+                self.vace_in_dim, self.dim, kernel_size=self.patch_size, stride=self.patch_size, device=device, dtype=torch.float32
+            )
+
+    def forward_orig(
+        self,
+        x,
+        t,
+        context,
+        vace_context,
+        clip_fea=None,
+        freqs=None,
+        transformer_options={},
+        **kwargs,
+    ):
+        # embeddings
+        x = self.patch_embedding(x.float()).to(x.dtype)
+        grid_sizes = x.shape[2:]
+        x = x.flatten(2).transpose(1, 2)
+
+        # time embeddings
+        e = self.time_embedding(
+            sinusoidal_embedding_1d(self.freq_dim, t).to(dtype=x[0].dtype))
+        e0 = self.time_projection(e).unflatten(1, (6, self.dim))
+
+        # context
+        context = self.text_embedding(context)
+
+        context_img_len = None
+        if clip_fea is not None:
+            if self.img_emb is not None:
+                context_clip = self.img_emb(clip_fea)  # bs x 257 x dim
+                context = torch.concat([context_clip, context], dim=1)
+            context_img_len = clip_fea.shape[-2]
+
+        c = self.vace_patch_embedding(vace_context.float()).to(vace_context.dtype)
+        c = c.flatten(2).transpose(1, 2)
+
+        # arguments
+        x_orig = x
+
+        patches_replace = transformer_options.get("patches_replace", {})
+        blocks_replace = patches_replace.get("dit", {})
+        for i, block in enumerate(self.blocks):
+            if ("double_block", i) in blocks_replace:
+                def block_wrap(args):
+                    out = {}
+                    out["img"] = block(args["img"], context=args["txt"], e=args["vec"], freqs=args["pe"], context_img_len=context_img_len)
+                    return out
+                out = blocks_replace[("double_block", i)]({"img": x, "txt": context, "vec": e0, "pe": freqs}, {"original_block": block_wrap})
+                x = out["img"]
+            else:
+                x = block(x, e=e0, freqs=freqs, context=context, context_img_len=context_img_len)
+
+            ii = self.vace_layers_mapping.get(i, None)
+            if ii is not None:
+                c_skip, c = self.vace_blocks[ii](c, x=x_orig, e=e0, freqs=freqs, context=context, context_img_len=context_img_len)
+            x += c_skip
+        # head
+        x = self.head(x, e)
+
+        # unpatchify
+        x = self.unpatchify(x, grid_sizes)
+        return x

comfy/model_base.py

@@ -1043,6 +1043,34 @@ class WAN21(BaseModel):
             out['clip_fea'] = comfy.conds.CONDRegular(clip_vision_output.penultimate_hidden_states)
         return out
 
+
+class WAN21_Vace(WAN21):
+    def __init__(self, model_config, model_type=ModelType.FLOW, image_to_video=False, device=None):
+        super(WAN21, self).__init__(model_config, model_type, device=device, unet_model=comfy.ldm.wan.model.VaceWanModel)
+        self.image_to_video = image_to_video
+
+    def extra_conds(self, **kwargs):
+        out = super().extra_conds(**kwargs)
+        noise = kwargs.get("noise", None)
+        noise_shape = list(noise.shape)
+        vace_frames = kwargs.get("vace_frames", None)
+        if vace_frames is None:
+            noise_shape[1] = 32
+            vace_frames = torch.zeros(noise_shape, device=noise.device, dtype=noise.dtype)
+
+        for i in range(0, vace_frames.shape[1], 16):
+            vace_frames = vace_frames.clone()
+            vace_frames[:, i:i + 16] = self.process_latent_in(vace_frames[:, i:i + 16])
+
+        mask = kwargs.get("vace_mask", None)
+        if mask is None:
+            noise_shape[1] = 64
+            mask = torch.ones(noise_shape, device=noise.device, dtype=noise.dtype)
+
+        out['vace_context'] = comfy.conds.CONDRegular(torch.cat([vace_frames.to(noise), mask.to(noise)], dim=1))
+        return out
+
+
 class Hunyuan3Dv2(BaseModel):
     def __init__(self, model_config, model_type=ModelType.FLOW, device=None):
         super().__init__(model_config, model_type, device=device, unet_model=comfy.ldm.hunyuan3d.model.Hunyuan3Dv2)

comfy/model_detection.py

@@ -317,10 +317,15 @@ def detect_unet_config(state_dict, key_prefix, metadata=None):
         dit_config["cross_attn_norm"] = True
         dit_config["eps"] = 1e-6
         dit_config["in_dim"] = state_dict['{}patch_embedding.weight'.format(key_prefix)].shape[1]
-        if '{}img_emb.proj.0.bias'.format(key_prefix) in state_dict_keys:
-            dit_config["model_type"] = "i2v"
+        if '{}vace_patch_embedding.weight'.format(key_prefix) in state_dict_keys:
+            dit_config["model_type"] = "vace"
+            dit_config["vace_in_dim"] = state_dict['{}vace_patch_embedding.weight'.format(key_prefix)].shape[1]
+            dit_config["vace_layers"] = count_blocks(state_dict_keys, '{}vace_blocks.'.format(key_prefix) + '{}.')
         else:
-            dit_config["model_type"] = "t2v"
+            if '{}img_emb.proj.0.bias'.format(key_prefix) in state_dict_keys:
+                dit_config["model_type"] = "i2v"
+            else:
+                dit_config["model_type"] = "t2v"
         flf_weight = state_dict.get('{}img_emb.emb_pos'.format(key_prefix))
         if flf_weight is not None:
             dit_config["flf_pos_embed_token_number"] = flf_weight.shape[1]

comfy/supported_models.py

@@ -987,6 +987,16 @@ class WAN21_FunControl2V(WAN21_T2V):
         out = model_base.WAN21(self, image_to_video=False, device=device)
         return out
 
+class WAN21_Vace(WAN21_T2V):
+    unet_config = {
+        "image_model": "wan2.1",
+        "model_type": "vace",
+    }
+
+    def get_model(self, state_dict, prefix="", device=None):
+        out = model_base.WAN21_Vace(self, image_to_video=False, device=device)
+        return out
+
 class Hunyuan3Dv2(supported_models_base.BASE):
     unet_config = {
         "image_model": "hunyuan3d2",
@@ -1055,6 +1065,6 @@ class HiDream(supported_models_base.BASE):
         return None #  TODO
 
 
-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, HunyuanVideoSkyreelsI2V, HunyuanVideoI2V, HunyuanVideo, CosmosT2V, CosmosI2V, Lumina2, WAN21_T2V, WAN21_I2V, WAN21_FunControl2V, Hunyuan3Dv2mini, Hunyuan3Dv2, HiDream]
+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, HunyuanVideoSkyreelsI2V, HunyuanVideoI2V, HunyuanVideo, CosmosT2V, CosmosI2V, Lumina2, WAN21_T2V, WAN21_I2V, WAN21_FunControl2V, WAN21_Vace, Hunyuan3Dv2mini, Hunyuan3Dv2, HiDream]
 
 models += [SVD_img2vid]

comfy_extras/nodes_wan.py

@@ -193,9 +193,115 @@ class WanFunInpaintToVideo:
         return flfv.encode(positive, negative, vae, width, height, length, batch_size, start_image=start_image, end_image=end_image, clip_vision_start_image=clip_vision_output)
 
 
+class WanVaceToVideo:
+    @classmethod
+    def INPUT_TYPES(s):
+        return {"required": {"positive": ("CONDITIONING", ),
+                             "negative": ("CONDITIONING", ),
+                             "vae": ("VAE", ),
+                             "width": ("INT", {"default": 832, "min": 16, "max": nodes.MAX_RESOLUTION, "step": 16}),
+                             "height": ("INT", {"default": 480, "min": 16, "max": nodes.MAX_RESOLUTION, "step": 16}),
+                             "length": ("INT", {"default": 81, "min": 1, "max": nodes.MAX_RESOLUTION, "step": 4}),
+                             "batch_size": ("INT", {"default": 1, "min": 1, "max": 4096}),
+                },
+                "optional": {"control_video": ("IMAGE", ),
+                             "control_masks": ("MASK", ),
+                             "reference_image": ("IMAGE", ),
+                }}
+
+    RETURN_TYPES = ("CONDITIONING", "CONDITIONING", "LATENT", "INT")
+    RETURN_NAMES = ("positive", "negative", "latent", "trim_latent")
+    FUNCTION = "encode"
+
+    CATEGORY = "conditioning/video_models"
+
+    EXPERIMENTAL = True
+
+    def encode(self, positive, negative, vae, width, height, length, batch_size, control_video=None, control_masks=None, reference_image=None):
+        latent_length = ((length - 1) // 4) + 1
+        if control_video is not None:
+            control_video = comfy.utils.common_upscale(control_video[:length].movedim(-1, 1), width, height, "bilinear", "center").movedim(1, -1)
+            if control_video.shape[0] < length:
+                control_video = torch.nn.functional.pad(control_video, (0, 0, 0, 0, 0, 0, 0, length - control_video.shape[0]), value=0.5)
+        else:
+            control_video = torch.ones((length, height, width, 3)) * 0.5
+
+        if reference_image is not None:
+            reference_image = comfy.utils.common_upscale(reference_image[:1].movedim(-1, 1), width, height, "bilinear", "center").movedim(1, -1)
+            reference_image = vae.encode(reference_image[:, :, :, :3])
+            reference_image = torch.cat([reference_image, comfy.latent_formats.Wan21().process_out(torch.zeros_like(reference_image))], dim=1)
+
+        if control_masks is None:
+            mask = torch.ones((length, height, width, 1))
+        else:
+            mask = control_masks
+            if mask.ndim == 3:
+                mask = mask.unsqueeze(1)
+            mask = comfy.utils.common_upscale(mask[:length], width, height, "bilinear", "center").movedim(1, -1)
+            if mask.shape[0] < length:
+                mask = torch.nn.functional.pad(mask, (0, 0, 0, 0, 0, 0, 0, length - mask.shape[0]), value=1.0)
+
+        control_video = control_video - 0.5
+        inactive = (control_video * (1 - mask)) + 0.5
+        reactive = (control_video * mask) + 0.5
+
+        inactive = vae.encode(inactive[:, :, :, :3])
+        reactive = vae.encode(reactive[:, :, :, :3])
+        control_video_latent = torch.cat((inactive, reactive), dim=1)
+        if reference_image is not None:
+            control_video_latent = torch.cat((reference_image, control_video_latent), dim=2)
+
+        vae_stride = 8
+        height_mask = height // vae_stride
+        width_mask = width // vae_stride
+        mask = mask.view(length, height_mask, vae_stride, width_mask, vae_stride)
+        mask = mask.permute(2, 4, 0, 1, 3)
+        mask = mask.reshape(vae_stride * vae_stride, length, height_mask, width_mask)
+        mask = torch.nn.functional.interpolate(mask.unsqueeze(0), size=(latent_length, height_mask, width_mask), mode='nearest-exact').squeeze(0)
+
+        trim_latent = 0
+        if reference_image is not None:
+            mask_pad = torch.zeros_like(mask[:, :reference_image.shape[2], :, :])
+            mask = torch.cat((mask_pad, mask), dim=1)
+            latent_length += reference_image.shape[2]
+            trim_latent = reference_image.shape[2]
+
+        mask = mask.unsqueeze(0)
+        positive = node_helpers.conditioning_set_values(positive, {"vace_frames": control_video_latent, "vace_mask": mask})
+        negative = node_helpers.conditioning_set_values(negative, {"vace_frames": control_video_latent, "vace_mask": mask})
+
+        latent = torch.zeros([batch_size, 16, latent_length, height // 8, width // 8], device=comfy.model_management.intermediate_device())
+        out_latent = {}
+        out_latent["samples"] = latent
+        return (positive, negative, out_latent, trim_latent)
+
+class TrimVideoLatent:
+    @classmethod
+    def INPUT_TYPES(s):
+        return {"required": { "samples": ("LATENT",),
+                              "trim_amount": ("INT", {"default": 0, "min": 0, "max": 99999}),
+                             }}
+
+    RETURN_TYPES = ("LATENT",)
+    FUNCTION = "op"
+
+    CATEGORY = "latent/video"
+
+    EXPERIMENTAL = True
+
+    def op(self, samples, trim_amount):
+        samples_out = samples.copy()
+
+        s1 = samples["samples"]
+        samples_out["samples"] = s1[:, :, trim_amount:]
+        return (samples_out,)
+
+
 NODE_CLASS_MAPPINGS = {
     "WanImageToVideo": WanImageToVideo,
     "WanFunControlToVideo": WanFunControlToVideo,
     "WanFunInpaintToVideo": WanFunInpaintToVideo,
     "WanFirstLastFrameToVideo": WanFirstLastFrameToVideo,
+    "WanVaceToVideo": WanVaceToVideo,
+    "TrimVideoLatent": TrimVideoLatent,
 }