15fpx

OmniAvatar/utils/audio_preprocess.py

@@ -18,4 +18,45 @@ def add_silence_to_audio_ffmpeg(audio_path, tmp_audio_path, silence_duration_s=0
     try:
         subprocess.run(cmd, check=True, capture_output=True, text=True)
     except subprocess.CalledProcessError as e:
-        raise RuntimeError(f"ffmpeg failed ({e.returncode}): {e.stderr.strip()}")
+        raise RuntimeError(f"ffmpeg failed ({e.returncode}): {e.stderr.strip()}")
+    
+def convert_video_to_15fps_ffmpeg(video_path, output_path=None):
+    """
+    Convert a video to 15 FPS using ffmpeg.
+
+    Parameters
+    ----------
+    video_path : str
+        Path to the input .mp4 video.
+    output_path : str, optional
+        Path for the output video. If None, a new file will be created next to the input.
+
+    Returns
+    -------
+    str
+        The output video path.
+    """
+
+    if not os.path.exists(video_path):
+        raise FileNotFoundError(f"Input video not found: {video_path}")
+
+    # Auto-generate output path if not provided
+    if output_path is None:
+        base, ext = os.path.splitext(video_path)
+        output_path = base + "_15fps.mp4"
+
+    cmd = [
+        "ffmpeg",
+        "-i", video_path,
+        "-filter:v", "fps=15",  # Set frame rate
+        "-c:a", "copy",         # Copy audio without re-encoding
+        "-y", output_path,      # Overwrite output
+        "-loglevel", "error"
+    ]
+
+    try:
+        subprocess.run(cmd, check=True, capture_output=True, text=True)
+    except subprocess.CalledProcessError as e:
+        raise RuntimeError(f"ffmpeg failed ({e.returncode}): {e.stderr.strip()}")
+
+    return output_path

app.py

@@ -81,7 +81,7 @@ import torchvision.transforms as TT
 from transformers import Wav2Vec2FeatureExtractor
 import torchvision.transforms as transforms
 import torch.nn.functional as F
-from OmniAvatar.utils.audio_preprocess import add_silence_to_audio_ffmpeg
+from OmniAvatar.utils.audio_preprocess import add_silence_to_audio_ffmpeg, convert_video_to_15fps_ffmpeg
 
 from diffusers import FluxKontextPipeline
 from diffusers.utils import load_image
@@ -715,8 +715,10 @@ def infer(image_path, audio_path, text, num_steps, session_id = None, progress=g
                             prompt_path = prompt_path,
                             audio_path=tmp2_audio_path if args.use_audio else None, 
                             prefix=f'result')
+    
+    video_path_15fps = convert_video_to_15fps_ffmpeg(video_paths[0])
 
-    return video_paths[0]
+    return video_path_15fps
 
 def apply_image(request):
     print('image applied')