Merge origin/dev into feat/seedvr2-comfyui-node

bridge/translator/inference.py

@@ -36,6 +36,7 @@ WRAPPER_TO_LIGHTX2V_FIELDS: Dict[str, str] = {
     "fixed_area": "fixed_area",
     # Sekotalk-specific.
     "prev_frame_length": "prev_frame_length",
+    "f2v_process": "f2v_process",
     # Distillation.
     "denoising_step_list": "denoising_step_list",
     "use_31_block": "use_31_block",

config_builder.py

@@ -140,7 +140,7 @@ class InferenceConfigBuilder:
             steps_str = optional_params["denoising_steps"]
             if steps_str and steps_str.strip():
                 try:
-                    steps_list = [int(s.strip()) for s in steps_str.split(",")]
+                    steps_list = [float(s.strip()) for s in steps_str.split(",")]
                     config.denoising_step_list = steps_list
                     config.infer_steps = len(steps_list)
                 except ValueError:
@@ -153,6 +153,7 @@ class InferenceConfigBuilder:
             "segment_length",
             "prev_frame_length",
             "use_tiny_vae",
+            "f2v_process",
         ]:
             if param in optional_params:
                 setattr(config, param, optional_params[param])

data_models.py

@@ -53,7 +53,7 @@ class InferenceConfig:
     use_31_block: bool = True
 
     # Optional parameters
-    denoising_step_list: Optional[List[int]] = None
+    denoising_step_list: Optional[List[float]] = None
     resize_mode: str = "adaptive"
     fixed_area: str = "720p"
     segment_length: int = 81
@@ -65,6 +65,7 @@ class InferenceConfig:
     negative_prompt: str = ""
     image_path: Optional[str] = None
     audio_path: Optional[str] = None
+    f2v_process: bool = False
 
     def to_dict(self) -> Dict[str, Any]:
         """Convert to dictionary, excluding None values."""

nodes/config.py

@@ -171,6 +171,13 @@ class LightX2VInferenceConfig:
                         "tooltip": "Use lightweight VAE to accelerate decoding",
                     },
                 ),
+                "f2v_process": (
+                    "BOOLEAN",
+                    {
+                        "default": False,
+                        "tooltip": "Use frame-to-video processing for SekoTalk",
+                    },
+                ),
             },
         }
 
@@ -199,6 +206,7 @@ class LightX2VInferenceConfig:
         segment_length=81,
         prev_frame_length=5,
         use_tiny_vae=False,
+        f2v_process=False,
     ):
         """Create basic inference configuration."""
         builder = InferenceConfigBuilder()
@@ -222,6 +230,7 @@ class LightX2VInferenceConfig:
             segment_length=segment_length,
             prev_frame_length=prev_frame_length,
             use_tiny_vae=use_tiny_vae,
+            f2v_process=f2v_process,
         )
 
         return (config.to_dict(),)