Merge 76bceb2fe5 into c709fcf0e7

generate.py

@@ -299,7 +299,7 @@ def generate(args):
         elif args.prompt_extend_method == "local_qwen":
             prompt_expander = QwenPromptExpander(
                 model_name=args.prompt_extend_model,
-                is_vl="i2v" in args.task,
+                is_vl="i2v" in args.task or "flf2v" in args.task,
                 device=rank)
         else:
             raise NotImplementedError(

gradio/vace.py

@@ -210,7 +210,7 @@ class VACEInference:
                                                                          [src_mask],
                                                                          [src_ref_images],
                                                                          num_frames=num_frames,
-                                                                         image_size=SIZE_CONFIGS[f"{output_height}*{output_width}"],
+                                                                         image_size=SIZE_CONFIGS[f"{output_width}*{output_height}"],
                                                                          device=self.pipe.device)
         video = self.pipe.generate(
             prompt,

wan/vace.py

@@ -200,6 +200,7 @@ class WanVace(WanT2V):
         else:
             raise NotImplementedError(f'image_size {image_size} is not supported')
 
+        image_size = (image_size[1], image_size[0])
         image_sizes = []
         for i, (sub_src_video, sub_src_mask) in enumerate(zip(src_video, src_mask)):
             if sub_src_mask is not None and sub_src_video is not None: