Merge branch 'master' of https://github.com/comfyanonymous/ComfyUI

2025-04-15 16:13:29 +00:00 · 2025-03-28 23:48:13 +09:00 · 2025-03-28 23:48:13 +09:00 · 311e110b1f
commit 311e110b1f
parent 3e70e821c5 2d17d8910c
18 changed files with 394 additions and 40 deletions
--- a/README.md
+++ b/README.md
@ -69,6 +69,8 @@ See what ComfyUI can do with the [example workflows](https://comfyanonymous.gith
   - [Hunyuan Video](https://comfyanonymous.github.io/ComfyUI_examples/hunyuan_video/)
   - [Nvidia Cosmos](https://comfyanonymous.github.io/ComfyUI_examples/cosmos/)
   - [Wan 2.1](https://comfyanonymous.github.io/ComfyUI_examples/wan/)
+- 3D Models
+   - [Hunyuan3D 2.0](https://docs.comfy.org/tutorials/3d/hunyuan3D-2)
 - [Stable Audio](https://comfyanonymous.github.io/ComfyUI_examples/audio/)
 - Asynchronous Queue system
 - Many optimizations: Only re-executes the parts of the workflow that changes between executions.
--- a/app/frontend_management.py
+++ b/app/frontend_management.py
@ -22,13 +22,21 @@ import app.logger
 # The path to the requirements.txt file
 req_path = Path(__file__).parents[1] / "requirements.txt"

+
 def frontend_install_warning_message():
    """The warning message to display when the frontend version is not up to date."""

    extra = ""
    if sys.flags.no_user_site:
        extra = "-s "
-    return f"Please install the updated requirements.txt file by running:\n{sys.executable} {extra}-m pip install -r {req_path}\n\nThis error is happening because the ComfyUI frontend is no longer shipped as part of the main repo but as a pip package instead.\n\nIf you are on the portable package you can run: update\\update_comfyui.bat to solve this problem"
+    return f"""
+Please install the updated requirements.txt file by running:
+{sys.executable} {extra}-m pip install -r {req_path}
+
+This error is happening because the ComfyUI frontend is no longer shipped as part of the main repo but as a pip package instead.
+
+If you are on the portable package you can run: update\\update_comfyui.bat to solve this problem
+""".strip()


 def check_frontend_version():
@ -43,7 +51,17 @@ def check_frontend_version():
        with open(req_path, "r", encoding="utf-8") as f:
            required_frontend = parse_version(f.readline().split("=")[-1])
        if frontend_version < required_frontend:
-            app.logger.log_startup_warning("________________________________________________________________________\nWARNING WARNING WARNING WARNING WARNING\n\nInstalled frontend version {} is lower than the recommended version {}.\n\n{}\n________________________________________________________________________".format('.'.join(map(str, frontend_version)), '.'.join(map(str, required_frontend)), frontend_install_warning_message()))
+            app.logger.log_startup_warning(
+                f"""
+________________________________________________________________________
+WARNING WARNING WARNING WARNING WARNING
+
+Installed frontend version {".".join(map(str, frontend_version))} is lower than the recommended version {".".join(map(str, required_frontend))}.
+
+{frontend_install_warning_message()}
+________________________________________________________________________
+""".strip()
+            )
        else:
            logging.info("ComfyUI frontend version: {}".format(frontend_version_str))
    except Exception as e:
@ -150,9 +168,20 @@ class FrontendManager:
    def default_frontend_path(cls) -> str:
        try:
            import comfyui_frontend_package
+
            return str(importlib.resources.files(comfyui_frontend_package) / "static")
        except ImportError:
-            logging.error(f"\n\n********** ERROR ***********\n\ncomfyui-frontend-package is not installed. {frontend_install_warning_message()}\n********** ERROR **********\n")
+            logging.error(
+                f"""
+********** ERROR ***********
+
+comfyui-frontend-package is not installed.
+
+{frontend_install_warning_message()}
+
+********** ERROR ***********
+""".strip()
+            )
            sys.exit(-1)

    @classmethod
@ -175,7 +204,9 @@ class FrontendManager:
        return match_result.group(1), match_result.group(2), match_result.group(3)

    @classmethod
-    def init_frontend_unsafe(cls, version_string: str, provider: Optional[FrontEndProvider] = None) -> str:
+    def init_frontend_unsafe(
+        cls, version_string: str, provider: Optional[FrontEndProvider] = None
+    ) -> str:
        """
        Initializes the frontend for the specified version.

@ -197,12 +228,20 @@ class FrontendManager:
        repo_owner, repo_name, version = cls.parse_version_string(version_string)

        if version.startswith("v"):
-            expected_path = str(Path(cls.CUSTOM_FRONTENDS_ROOT) / f"{repo_owner}_{repo_name}" / version.lstrip("v"))
+            expected_path = str(
+                Path(cls.CUSTOM_FRONTENDS_ROOT)
+                / f"{repo_owner}_{repo_name}"
+                / version.lstrip("v")
+            )
            if os.path.exists(expected_path):
-                logging.info(f"Using existing copy of specific frontend version tag: {repo_owner}/{repo_name}@{version}")
+                logging.info(
+                    f"Using existing copy of specific frontend version tag: {repo_owner}/{repo_name}@{version}"
+                )
                return expected_path

-        logging.info(f"Initializing frontend: {repo_owner}/{repo_name}@{version}, requesting version details from GitHub...")
+        logging.info(
+            f"Initializing frontend: {repo_owner}/{repo_name}@{version}, requesting version details from GitHub..."
+        )

        provider = provider or FrontEndProvider(repo_owner, repo_name)
        release = provider.get_release(version)
--- a/comfy/ldm/modules/attention.py
+++ b/comfy/ldm/modules/attention.py
@ -471,7 +471,7 @@ def attention_pytorch(q, k, v, heads, mask=None, attn_precision=None, skip_resha
 def attention_sage(q, k, v, heads, mask=None, attn_precision=None, skip_reshape=False, skip_output_reshape=False):
    if skip_reshape:
        b, _, _, dim_head = q.shape
-        tensor_layout="HND"
+        tensor_layout = "HND"
    else:
        b, _, dim_head = q.shape
        dim_head //= heads
@ -479,7 +479,7 @@ def attention_sage(q, k, v, heads, mask=None, attn_precision=None, skip_reshape=
            lambda t: t.view(b, -1, heads, dim_head),
            (q, k, v),
        )
-        tensor_layout="NHD"
+        tensor_layout = "NHD"

    if mask is not None:
        # add a batch dimension if there isn't already one
@ -489,7 +489,17 @@ def attention_sage(q, k, v, heads, mask=None, attn_precision=None, skip_reshape=
        if mask.ndim == 3:
            mask = mask.unsqueeze(1)

-    out = sageattn(q, k, v, attn_mask=mask, is_causal=False, tensor_layout=tensor_layout)
+    try:
+        out = sageattn(q, k, v, attn_mask=mask, is_causal=False, tensor_layout=tensor_layout)
+    except Exception as e:
+        logging.error("Error running sage attention: {}, using pytorch attention instead.".format(e))
+        if tensor_layout == "NHD":
+            q, k, v = map(
+                lambda t: t.transpose(1, 2),
+                (q, k, v),
+            )
+        return attention_pytorch(q, k, v, heads, mask=mask, skip_reshape=True, skip_output_reshape=skip_output_reshape)
+
    if tensor_layout == "HND":
        if not skip_output_reshape:
            out = (
--- a/comfy/model_base.py
+++ b/comfy/model_base.py
@ -59,6 +59,7 @@ class ModelType(Enum):
    FLOW = 6
    V_PREDICTION_CONTINUOUS = 7
    FLUX = 8
+    IMG_TO_IMG = 9


 from comfy.model_sampling import EPS, V_PREDICTION, EDM, ModelSamplingDiscrete, ModelSamplingContinuousEDM, StableCascadeSampling, ModelSamplingContinuousV
@ -89,6 +90,8 @@ def model_sampling(model_config, model_type):
    elif model_type == ModelType.FLUX:
        c = comfy.model_sampling.CONST
        s = comfy.model_sampling.ModelSamplingFlux
+    elif model_type == ModelType.IMG_TO_IMG:
+        c = comfy.model_sampling.IMG_TO_IMG

    class ModelSampling(s, c):
        pass
@ -140,6 +143,7 @@ class BaseModel(torch.nn.Module):
    def _apply_model(self, x, t, c_concat=None, c_crossattn=None, control=None, transformer_options={}, **kwargs):
        sigma = t
        xc = self.model_sampling.calculate_input(sigma, x)
+
        if c_concat is not None:
            xc = torch.cat([xc] + [c_concat], dim=1)

@ -601,6 +605,19 @@ class SDXL_instructpix2pix(IP2P, SDXL):
        else:
            self.process_ip2p_image_in = lambda image: image #diffusers ip2p

+class Lotus(BaseModel):
+    def extra_conds(self, **kwargs):
+        out = {}
+        cross_attn = kwargs.get("cross_attn", None)
+        out['c_crossattn'] = comfy.conds.CONDCrossAttn(cross_attn)
+        device = kwargs["device"]
+        task_emb = torch.tensor([1, 0]).float().to(device)
+        task_emb = torch.cat([torch.sin(task_emb), torch.cos(task_emb)]).unsqueeze(0)
+        out['y'] = comfy.conds.CONDRegular(task_emb)
+        return out
+
+    def __init__(self, model_config, model_type=ModelType.IMG_TO_IMG, device=None):
+        super().__init__(model_config, model_type, device=device)

 class StableCascade_C(BaseModel):
    def __init__(self, model_config, model_type=ModelType.STABLE_CASCADE, device=None):
@ -975,7 +992,8 @@ class WAN21(BaseModel):

    def concat_cond(self, **kwargs):
        noise = kwargs.get("noise", None)
-        if self.diffusion_model.patch_embedding.weight.shape[1] == noise.shape[1]:
+        extra_channels = self.diffusion_model.patch_embedding.weight.shape[1] - noise.shape[1]
+        if extra_channels == 0:
            return None

        image = kwargs.get("concat_latent_image", None)
@ -983,23 +1001,33 @@ class WAN21(BaseModel):

        if image is None:
            image = torch.zeros_like(noise)
+            shape_image = list(noise.shape)
+            shape_image[1] = extra_channels
+            image = torch.zeros(shape_image, dtype=noise.dtype, layout=noise.layout, device=noise.device)
+        else:
+            image = utils.common_upscale(image.to(device), noise.shape[-1], noise.shape[-2], "bilinear", "center")
+            for i in range(0, image.shape[1], 16):
+                image[:, i: i + 16] = self.process_latent_in(image[:, i: i + 16])
+            image = utils.resize_to_batch_size(image, noise.shape[0])

-        image = utils.common_upscale(image.to(device), noise.shape[-1], noise.shape[-2], "bilinear", "center")
-        image = self.process_latent_in(image)
-        image = utils.resize_to_batch_size(image, noise.shape[0])
-
-        if not self.image_to_video:
+        if not self.image_to_video or extra_channels == image.shape[1]:
            return image

+        if image.shape[1] > (extra_channels - 4):
+            image = image[:, :(extra_channels - 4)]
+
        mask = kwargs.get("concat_mask", kwargs.get("denoise_mask", None))
        if mask is None:
            mask = torch.zeros_like(noise)[:, :4]
        else:
-            mask = 1.0 - torch.mean(mask, dim=1, keepdim=True)
+            if mask.shape[1] != 4:
+                mask = torch.mean(mask, dim=1, keepdim=True)
+            mask = 1.0 - mask
            mask = utils.common_upscale(mask.to(device), noise.shape[-1], noise.shape[-2], "bilinear", "center")
            if mask.shape[-3] < noise.shape[-3]:
                mask = torch.nn.functional.pad(mask, (0, 0, 0, 0, 0, noise.shape[-3] - mask.shape[-3]), mode='constant', value=0)
-            mask = mask.repeat(1, 4, 1, 1, 1)
+            if mask.shape[1] == 1:
+                mask = mask.repeat(1, 4, 1, 1, 1)
            mask = utils.resize_to_batch_size(mask, noise.shape[0])

        return torch.cat((mask, image), dim=1)
--- a/comfy/model_detection.py
+++ b/comfy/model_detection.py
@ -682,8 +682,13 @@ def unet_config_from_diffusers_unet(state_dict, dtype=None):
            'transformer_depth_output': [1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0],
            'use_temporal_attention': False, 'use_temporal_resblock': False}

+    LotusD = {'use_checkpoint': False, 'image_size': 32, 'out_channels': 4, 'use_spatial_transformer': True, 'legacy': False, 'adm_in_channels': 4,
+            'dtype': dtype, 'in_channels': 4, 'model_channels': 320, 'num_res_blocks': [2, 2, 2, 2], 'transformer_depth': [1, 1, 1, 1, 1, 1, 0, 0],
+            'channel_mult': [1, 2, 4, 4], 'transformer_depth_middle': 1, 'use_linear_in_transformer': True, 'context_dim': 1024, 'num_heads': 8,
+            'transformer_depth_output': [1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0],
+            'use_temporal_attention': False, 'use_temporal_resblock': False}

-    supported_models = [SDXL, SDXL_refiner, SD21, SD15, SD21_uncliph, SD21_unclipl, SDXL_mid_cnet, SDXL_small_cnet, SDXL_diffusers_inpaint, SSD_1B, Segmind_Vega, KOALA_700M, KOALA_1B, SD09_XS, SD_XS, SDXL_diffusers_ip2p, SD15_diffusers_inpaint]
+    supported_models = [LotusD, SDXL, SDXL_refiner, SD21, SD15, SD21_uncliph, SD21_unclipl, SDXL_mid_cnet, SDXL_small_cnet, SDXL_diffusers_inpaint, SSD_1B, Segmind_Vega, KOALA_700M, KOALA_1B, SD09_XS, SD_XS, SDXL_diffusers_ip2p, SD15_diffusers_inpaint]

    for unet_config in supported_models:
        matches = True
--- a/comfy/model_management.py
+++ b/comfy/model_management.py
@ -46,6 +46,32 @@ cpu_state = CPUState.GPU

 total_vram = 0

+def get_supported_float8_types():
+    float8_types = []
+    try:
+        float8_types.append(torch.float8_e4m3fn)
+    except:
+        pass
+    try:
+        float8_types.append(torch.float8_e4m3fnuz)
+    except:
+        pass
+    try:
+        float8_types.append(torch.float8_e5m2)
+    except:
+        pass
+    try:
+        float8_types.append(torch.float8_e5m2fnuz)
+    except:
+        pass
+    try:
+        float8_types.append(torch.float8_e8m0fnu)
+    except:
+        pass
+    return float8_types
+
+FLOAT8_TYPES = get_supported_float8_types()
+
 xpu_available = False
 torch_version = ""
 try:
@ -701,11 +727,8 @@ def unet_dtype(device=None, model_params=0, supported_dtypes=[torch.float16, tor
        return torch.float8_e5m2

    fp8_dtype = None
-    try:
-        if weight_dtype in [torch.float8_e4m3fn, torch.float8_e5m2]:
-            fp8_dtype = weight_dtype
-    except:
-        pass
+    if weight_dtype in FLOAT8_TYPES:
+        fp8_dtype = weight_dtype

    if fp8_dtype is not None:
        if supports_fp8_compute(device): #if fp8 compute is supported the casting is most likely not expensive
--- a/comfy/model_sampling.py
+++ b/comfy/model_sampling.py
@ -69,6 +69,15 @@ class CONST:
        sigma = sigma.view(sigma.shape[:1] + (1,) * (latent.ndim - 1))
        return latent / (1.0 - sigma)

+class X0(EPS):
+    def calculate_denoised(self, sigma, model_output, model_input):
+        return model_output
+
+class IMG_TO_IMG(X0):
+    def calculate_input(self, sigma, noise):
+        return noise
+
+
 class ModelSamplingDiscrete(torch.nn.Module):
    def __init__(self, model_config=None, zsnr=None):
        super().__init__()
--- a/comfy/supported_models.py
+++ b/comfy/supported_models.py
@ -506,6 +506,22 @@ class SDXL_instructpix2pix(SDXL):
    def get_model(self, state_dict, prefix="", device=None):
        return model_base.SDXL_instructpix2pix(self, model_type=self.model_type(state_dict, prefix), device=device)

+class LotusD(SD20):
+    unet_config = {
+        "model_channels": 320,
+        "use_linear_in_transformer": True,
+        "use_temporal_attention": False,
+        "adm_in_channels": 4,
+        "in_channels": 4,
+    }
+
+    unet_extra_config = {
+        "num_classes": 'sequential'
+    }
+
+    def get_model(self, state_dict, prefix="", device=None):
+        return model_base.Lotus(self, device=device)
+
 class SD3(supported_models_base.BASE):
    unet_config = {
        "in_channels": 16,
@ -953,12 +969,24 @@ class WAN21_I2V(WAN21_T2V):
    unet_config = {
        "image_model": "wan2.1",
        "model_type": "i2v",
+        "in_dim": 36,
    }

    def get_model(self, state_dict, prefix="", device=None):
        out = model_base.WAN21(self, image_to_video=True, device=device)
        return out

+class WAN21_FunControl2V(WAN21_T2V):
+    unet_config = {
+        "image_model": "wan2.1",
+        "model_type": "i2v",
+        "in_dim": 48,
+    }
+
+    def get_model(self, state_dict, prefix="", device=None):
+        out = model_base.WAN21(self, image_to_video=False, device=device)
+        return out
+
 class Hunyuan3Dv2(supported_models_base.BASE):
    unet_config = {
        "image_model": "hunyuan3d2",
@ -997,6 +1025,6 @@ class Hunyuan3Dv2mini(Hunyuan3Dv2):

    latent_format = latent_formats.Hunyuan3Dv2mini

-models = [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, Hunyuan3Dv2mini, Hunyuan3Dv2]
+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]

 models += [SVD_img2vid]
--- a/comfy_extras/nodes_cfg.py
+++ b/comfy_extras/nodes_cfg.py
@ -0,0 +1,45 @@
+import torch
+
+# https://github.com/WeichenFan/CFG-Zero-star
+def optimized_scale(positive, negative):
+    positive_flat = positive.reshape(positive.shape[0], -1)
+    negative_flat = negative.reshape(negative.shape[0], -1)
+
+    # Calculate dot production
+    dot_product = torch.sum(positive_flat * negative_flat, dim=1, keepdim=True)
+
+    # Squared norm of uncondition
+    squared_norm = torch.sum(negative_flat ** 2, dim=1, keepdim=True) + 1e-8
+
+    # st_star = v_cond^T * v_uncond / ||v_uncond||^2
+    st_star = dot_product / squared_norm
+
+    return st_star.reshape([positive.shape[0]] + [1] * (positive.ndim - 1))
+
+class CFGZeroStar:
+    @classmethod
+    def INPUT_TYPES(s):
+        return {"required": {"model": ("MODEL",),
+                            }}
+    RETURN_TYPES = ("MODEL",)
+    RETURN_NAMES = ("patched_model",)
+    FUNCTION = "patch"
+    CATEGORY = "advanced/guidance"
+
+    def patch(self, model):
+        m = model.clone()
+        def cfg_zero_star(args):
+            guidance_scale = args['cond_scale']
+            x = args['input']
+            cond_p = args['cond_denoised']
+            uncond_p = args['uncond_denoised']
+            out = args["denoised"]
+            alpha = optimized_scale(x - cond_p, x - uncond_p)
+
+            return out + uncond_p * (alpha - 1.0)  + guidance_scale * uncond_p * (1.0 - alpha)
+        m.set_model_sampler_post_cfg_function(cfg_zero_star)
+        return (m, )
+
+NODE_CLASS_MAPPINGS = {
+    "CFGZeroStar": CFGZeroStar
+}
--- a/comfy_extras/nodes_load_3d.py
+++ b/comfy_extras/nodes_load_3d.py
@ -21,8 +21,8 @@ class Load3D():
            "height": ("INT", {"default": 1024, "min": 1, "max": 4096, "step": 1}),
        }}

-    RETURN_TYPES = ("IMAGE", "MASK", "STRING")
-    RETURN_NAMES = ("image", "mask", "mesh_path")
+    RETURN_TYPES = ("IMAGE", "MASK", "STRING", "IMAGE", "IMAGE")
+    RETURN_NAMES = ("image", "mask", "mesh_path", "normal", "lineart")

    FUNCTION = "process"
    EXPERIMENTAL = True
@ -32,12 +32,16 @@ class Load3D():
    def process(self, model_file, image, **kwargs):
        image_path = folder_paths.get_annotated_filepath(image['image'])
        mask_path = folder_paths.get_annotated_filepath(image['mask'])
+        normal_path = folder_paths.get_annotated_filepath(image['normal'])
+        lineart_path = folder_paths.get_annotated_filepath(image['lineart'])

        load_image_node = nodes.LoadImage()
        output_image, ignore_mask = load_image_node.load_image(image=image_path)
        ignore_image, output_mask = load_image_node.load_image(image=mask_path)
+        normal_image, ignore_mask2 = load_image_node.load_image(image=normal_path)
+        lineart_image, ignore_mask3 = load_image_node.load_image(image=lineart_path)

-        return output_image, output_mask, model_file,
+        return output_image, output_mask, model_file, normal_image, lineart_image

 class Load3DAnimation():
    @classmethod
@ -55,8 +59,8 @@ class Load3DAnimation():
            "height": ("INT", {"default": 1024, "min": 1, "max": 4096, "step": 1}),
        }}

-    RETURN_TYPES = ("IMAGE", "MASK", "STRING")
-    RETURN_NAMES = ("image", "mask", "mesh_path")
+    RETURN_TYPES = ("IMAGE", "MASK", "STRING", "IMAGE")
+    RETURN_NAMES = ("image", "mask", "mesh_path", "normal")

    FUNCTION = "process"
    EXPERIMENTAL = True
@ -66,12 +70,14 @@ class Load3DAnimation():
    def process(self, model_file, image, **kwargs):
        image_path = folder_paths.get_annotated_filepath(image['image'])
        mask_path = folder_paths.get_annotated_filepath(image['mask'])
+        normal_path = folder_paths.get_annotated_filepath(image['normal'])

        load_image_node = nodes.LoadImage()
        output_image, ignore_mask = load_image_node.load_image(image=image_path)
        ignore_image, output_mask = load_image_node.load_image(image=mask_path)
+        normal_image, ignore_mask2 = load_image_node.load_image(image=normal_path)

-        return output_image, output_mask, model_file,
+        return output_image, output_mask, model_file, normal_image

 class Preview3D():
    @classmethod
--- a/comfy_extras/nodes_lotus.py
+++ b/comfy_extras/nodes_lotus.py
--- a/comfy_extras/nodes_model_advanced.py
+++ b/comfy_extras/nodes_model_advanced.py
@ -20,10 +20,6 @@ class LCM(comfy.model_sampling.EPS):

        return c_out * x0 + c_skip * model_input

-class X0(comfy.model_sampling.EPS):
-    def calculate_denoised(self, sigma, model_output, model_input):
-        return model_output
-
 class ModelSamplingDiscreteDistilled(comfy.model_sampling.ModelSamplingDiscrete):
    original_timesteps = 50

@ -56,7 +52,7 @@ class ModelSamplingDiscrete:
    @classmethod
    def INPUT_TYPES(s):
        return {"required": { "model": ("MODEL",),
-                              "sampling": (["eps", "v_prediction", "lcm", "x0"],),
+                              "sampling": (["eps", "v_prediction", "lcm", "x0", "img_to_img"],),
                              "zsnr": ("BOOLEAN", {"default": False}),
                              }}

@ -77,7 +73,9 @@ class ModelSamplingDiscrete:
            sampling_type = LCM
            sampling_base = ModelSamplingDiscreteDistilled
        elif sampling == "x0":
-            sampling_type = X0
+            sampling_type = comfy.model_sampling.X0
+        elif sampling == "img_to_img":
+            sampling_type = comfy.model_sampling.IMG_TO_IMG

        class ModelSamplingAdvanced(sampling_base, sampling_type):
            pass
--- a/comfy_extras/nodes_model_merging_model_specific.py
+++ b/comfy_extras/nodes_model_merging_model_specific.py
@ -244,6 +244,30 @@ 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"
+    DESCRIPTION = "1.3B model has 30 blocks, 14B model has 40 blocks. Image to video model has the extra img_emb."
+
+    @classmethod
+    def INPUT_TYPES(s):
+        arg_dict = { "model1": ("MODEL",),
+                              "model2": ("MODEL",)}
+
+        argument = ("FLOAT", {"default": 1.0, "min": 0.0, "max": 1.0, "step": 0.01})
+
+        arg_dict["patch_embedding."] = argument
+        arg_dict["time_embedding."] = argument
+        arg_dict["time_projection."] = argument
+        arg_dict["text_embedding."] = argument
+        arg_dict["img_emb."] = argument
+
+        for i in range(40):
+            arg_dict["blocks.{}.".format(i)] = argument
+
+        arg_dict["head."] = argument
+
+        return {"required": arg_dict}
+
 NODE_CLASS_MAPPINGS = {
    "ModelMergeSD1": ModelMergeSD1,
    "ModelMergeSD2": ModelMergeSD1, #SD1 and SD2 have the same blocks
@ -256,4 +280,5 @@ NODE_CLASS_MAPPINGS = {
    "ModelMergeLTXV": ModelMergeLTXV,
    "ModelMergeCosmos7B": ModelMergeCosmos7B,
    "ModelMergeCosmos14B": ModelMergeCosmos14B,
+    "ModelMergeWAN2_1": ModelMergeWAN2_1,
 }
--- a/comfy_extras/nodes_wan.py
+++ b/comfy_extras/nodes_wan.py
@ -3,6 +3,7 @@ import node_helpers
 import torch
 import comfy.model_management
 import comfy.utils
+import comfy.latent_formats


 class WanImageToVideo:
@ -49,6 +50,110 @@ class WanImageToVideo:
        return (positive, negative, out_latent)


+class WanFunControlToVideo:
+    @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": {"clip_vision_output": ("CLIP_VISION_OUTPUT", ),
+                             "start_image": ("IMAGE", ),
+                             "control_video": ("IMAGE", ),
+                }}
+
+    RETURN_TYPES = ("CONDITIONING", "CONDITIONING", "LATENT")
+    RETURN_NAMES = ("positive", "negative", "latent")
+    FUNCTION = "encode"
+
+    CATEGORY = "conditioning/video_models"
+
+    def encode(self, positive, negative, vae, width, height, length, batch_size, start_image=None, clip_vision_output=None, control_video=None):
+        latent = torch.zeros([batch_size, 16, ((length - 1) // 4) + 1, height // 8, width // 8], device=comfy.model_management.intermediate_device())
+        concat_latent = torch.zeros([batch_size, 16, ((length - 1) // 4) + 1, height // 8, width // 8], device=comfy.model_management.intermediate_device())
+        concat_latent = comfy.latent_formats.Wan21().process_out(concat_latent)
+        concat_latent = concat_latent.repeat(1, 2, 1, 1, 1)
+
+        if start_image is not None:
+            start_image = comfy.utils.common_upscale(start_image[:length].movedim(-1, 1), width, height, "bilinear", "center").movedim(1, -1)
+            concat_latent_image = vae.encode(start_image[:, :, :, :3])
+            concat_latent[:,16:,:concat_latent_image.shape[2]] = concat_latent_image[:,:,:concat_latent.shape[2]]
+
+        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)
+            concat_latent_image = vae.encode(control_video[:, :, :, :3])
+            concat_latent[:,:16,:concat_latent_image.shape[2]] = concat_latent_image[:,:,:concat_latent.shape[2]]
+
+        positive = node_helpers.conditioning_set_values(positive, {"concat_latent_image": concat_latent})
+        negative = node_helpers.conditioning_set_values(negative, {"concat_latent_image": concat_latent})
+
+        if clip_vision_output is not None:
+            positive = node_helpers.conditioning_set_values(positive, {"clip_vision_output": clip_vision_output})
+            negative = node_helpers.conditioning_set_values(negative, {"clip_vision_output": clip_vision_output})
+
+        out_latent = {}
+        out_latent["samples"] = latent
+        return (positive, negative, out_latent)
+
+class WanFunInpaintToVideo:
+    @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": {"clip_vision_output": ("CLIP_VISION_OUTPUT", ),
+                             "start_image": ("IMAGE", ),
+                             "end_image": ("IMAGE", ),
+                }}
+
+    RETURN_TYPES = ("CONDITIONING", "CONDITIONING", "LATENT")
+    RETURN_NAMES = ("positive", "negative", "latent")
+    FUNCTION = "encode"
+
+    CATEGORY = "conditioning/video_models"
+
+    def encode(self, positive, negative, vae, width, height, length, batch_size, start_image=None, end_image=None, clip_vision_output=None):
+        latent = torch.zeros([batch_size, 16, ((length - 1) // 4) + 1, height // 8, width // 8], device=comfy.model_management.intermediate_device())
+        if start_image is not None:
+            start_image = comfy.utils.common_upscale(start_image[:length].movedim(-1, 1), width, height, "bilinear", "center").movedim(1, -1)
+        if end_image is not None:
+            end_image = comfy.utils.common_upscale(end_image[-length:].movedim(-1, 1), width, height, "bilinear", "center").movedim(1, -1)
+
+        image = torch.ones((length, height, width, 3)) * 0.5
+        mask = torch.ones((1, 1, latent.shape[2] * 4, latent.shape[-2], latent.shape[-1]))
+
+        if start_image is not None:
+            image[:start_image.shape[0]] = start_image
+            mask[:, :, :start_image.shape[0] + 3] = 0.0
+
+        if end_image is not None:
+            image[-end_image.shape[0]:] = end_image
+            mask[:, :, -end_image.shape[0]:] = 0.0
+
+        concat_latent_image = vae.encode(image[:, :, :, :3])
+        mask = mask.view(1, mask.shape[2] // 4, 4, mask.shape[3], mask.shape[4]).transpose(1, 2)
+        positive = node_helpers.conditioning_set_values(positive, {"concat_latent_image": concat_latent_image, "concat_mask": mask})
+        negative = node_helpers.conditioning_set_values(negative, {"concat_latent_image": concat_latent_image, "concat_mask": mask})
+
+        if clip_vision_output is not None:
+            positive = node_helpers.conditioning_set_values(positive, {"clip_vision_output": clip_vision_output})
+            negative = node_helpers.conditioning_set_values(negative, {"clip_vision_output": clip_vision_output})
+
+        out_latent = {}
+        out_latent["samples"] = latent
+        return (positive, negative, out_latent)
+
 NODE_CLASS_MAPPINGS = {
    "WanImageToVideo": WanImageToVideo,
+    "WanFunControlToVideo": WanFunControlToVideo,
+    "WanFunInpaintToVideo": WanFunInpaintToVideo,
 }
--- a/comfyui_version.py
+++ b/comfyui_version.py
@ -1,3 +1,3 @@
 # This file is automatically generated by the build process when version is
 # updated in pyproject.toml.
-__version__ = "0.3.26"
+__version__ = "0.3.27"
--- a/nodes.py
+++ b/nodes.py
@ -2310,8 +2310,10 @@ def init_builtin_extra_nodes():
        "nodes_video.py",
        "nodes_lumina2.py",
        "nodes_wan.py",
+        "nodes_lotus.py",
        "nodes_hunyuan3d.py",
        "nodes_primitive.py",
+        "nodes_cfg.py",
    ]

    import_failed = []
--- a/pyproject.toml
+++ b/pyproject.toml
@ -1,6 +1,6 @@
 [project]
 name = "ComfyUI"
-version = "0.3.26"
+version = "0.3.27"
 readme = "README.md"
 license = { file = "LICENSE" }
 requires-python = ">=3.9"
--- a/requirements.txt
+++ b/requirements.txt
@ -1,4 +1,4 @@
-comfyui-frontend-package==1.13.9
+comfyui-frontend-package==1.14.6
 torch
 torchsde
 torchvision