refactor: update DefaultRunner to return audio as None and modify...

refactor: update DefaultRunner to return audio as None and modify WanAudioRunner to handle audio saving and frame interpolation

lightx2v/models/runners/default_runner.py

@@ -240,4 +240,5 @@ class DefaultRunner(BaseRunner):
         torch.cuda.empty_cache()
         gc.collect()
 
-        return images
+        # Return (images, audio) - audio is None for default runner
+        return images, None

lightx2v/models/runners/wan/wan_audio_runner.py

@@ -432,7 +432,7 @@ class WanAudioRunner(WanRunner):
         ret["target_shape"] = self.config.target_shape
         return ret
 
-    def run(self):
+    def run(self, save_video=True):
         def load_audio(in_path: str, sr: float = 16000):
             audio_array, ori_sr = ta.load(in_path)
             audio_array = ta.functional.resample(audio_array.mean(0), orig_freq=ori_sr, new_freq=sr)
@@ -548,7 +548,7 @@ class WanAudioRunner(WanRunner):
                 self.model.scheduler.reset()
 
             if prev_latents is not None:
-                ltnt_channel, nframe, height, width = self.model.scheduler.latents.shape
+                _, nframe, height, width = self.model.scheduler.latents.shape
                 # bs = 1
                 frames_n = (nframe - 1) * 4 + 1
                 prev_frame_len = max((prev_len - 1) * 4 + 1, 0)
@@ -592,15 +592,39 @@ class WanAudioRunner(WanRunner):
 
         gen_lvideo = torch.cat(gen_video_list, dim=2).float()
         merge_audio = np.concatenate(cut_audio_list, axis=0).astype(np.float32)
+        comfyui_images = vae_to_comfyui_image(gen_lvideo)
+
+        # Apply frame interpolation if configured
+        if "video_frame_interpolation" in self.config:
+            assert self.vfi_model is not None and self.config["video_frame_interpolation"].get("target_fps", None) is not None
+            interpolation_target_fps = self.config["video_frame_interpolation"]["target_fps"]
+            logger.info(f"Interpolating frames from {target_fps} to {interpolation_target_fps}")
+            comfyui_images = self.vfi_model.interpolate_frames(
+                comfyui_images,
+                source_fps=target_fps,
+                target_fps=interpolation_target_fps,
+            )
+            # Update target_fps for saving
+            target_fps = interpolation_target_fps
+
+        # Convert audio to ComfyUI format
+        # Convert numpy array to torch tensor and add batch dimension
+        audio_waveform = torch.from_numpy(merge_audio).unsqueeze(0).unsqueeze(0)  # [batch, channels, samples]
+        comfyui_audio = {"waveform": audio_waveform, "sample_rate": audio_sr}
+
+        # Save video if requested
+        if save_video and self.config.get("save_video_path", None):
             out_path = os.path.join("./", "video_merge.mp4")
             audio_file = os.path.join("./", "audio_merge.wav")
-        comfyui_images = vae_to_comfyui_image(gen_lvideo)
+            # Use the updated target_fps (after interpolation if applied)
             save_to_video(comfyui_images, out_path, target_fps)
             save_audio(merge_audio, audio_file, out_path, output_path=self.config.get("save_video_path", None))
             os.remove(out_path)
             os.remove(audio_file)
 
-    def run_pipeline(self):
+        return comfyui_images, comfyui_audio
+
+    def run_pipeline(self, save_video=True):
         if self.config["use_prompt_enhancer"]:
             self.config["prompt_enhanced"] = self.post_prompt_enhancer()
 
@@ -609,8 +633,10 @@ class WanAudioRunner(WanRunner):
 
         self.init_scheduler()
         self.model.scheduler.prepare(self.inputs["image_encoder_output"])
-        self.run()
+        images, audio = self.run(save_video)  # run() now returns both images and audio
         self.end_run()
 
         gc.collect()
         torch.cuda.empty_cache()
+
+        return images, audio