Update FantasyTalking/infer.py

FantasyTalking/infer.py

@@ -176,8 +176,8 @@ def main(args, pipe, fantasytalking, wav2vec_processor, wav2vec):
     duration = librosa.get_duration(filename=args.audio_path)
     print(f"🎞️ Duration: {duration:.2f}s")
    
-    latents_num_frames = min(int(duration * args.fps / 4), args.max_num_frames // 4)
-    num_frames = (latents_num_frames - 1) * 4
+  
+    num_frames = min(int(args.fps * duration), args.max_num_frames)
     print(f"📽️ Calculated number of frames: {num_frames}")
 
     print("🎧 Extracting audio features...")