syncnet-python 0.2.1__py3-none-any.whl → 0.2.3__py3-none-any.whl

syncnet_python/__init__.py

@@ -4,23 +4,34 @@ This package provides a PyTorch implementation of SyncNet for detecting
 synchronization between audio and video in multimedia content.
 """
 
-__version__ = "0.1.1"
+__version__ = "0.2.3"
 
-# Import main components
+# Import main components. An ImportError here (for example a missing
+# dependency or an incompatible scenedetect version) is raised with its
+# original message instead of being replaced by None.
 try:
     from .syncnet_pipeline import SyncNetPipeline
     from .SyncNetModel import S as SyncNetModel
     from .SyncNetInstance import SyncNetInstance
-except ImportError:
-    # Fallback for development
-    SyncNetPipeline = None
-    SyncNetModel = None
-    SyncNetInstance = None
+    from .safe_syncnet_utils import (
+        safe_syncnet_inference,
+        extract_audio_from_video,
+        calculate_lse_metrics
+    )
+except ImportError as e:
+    raise ImportError(
+        f"syncnet_python could not import its components: {e}. "
+        "Check that the dependencies are installed with the versions listed "
+        "in pyproject.toml (scenedetect must be >=0.6,<0.7)."
+    ) from e
 
 __all__ = [
     "SyncNetPipeline",
     "SyncNetModel", 
     "SyncNetInstance",
+    "safe_syncnet_inference",
+    "extract_audio_from_video", 
+    "calculate_lse_metrics",
     "__version__"
 ]
 

syncnet_python/safe_syncnet_utils.py

@@ -0,0 +1,154 @@
+"""Safe SyncNet utilities with comprehensive error handling."""
+
+import logging
+import tempfile
+import os
+from pathlib import Path
+from typing import Optional, Tuple, List
+try:
+    from .syncnet_pipeline import SyncNetPipeline
+except ImportError:
+    from syncnet_pipeline import SyncNetPipeline
+
+
+def safe_syncnet_inference(
+    pipeline: SyncNetPipeline,
+    video_path: str,
+    audio_path: Optional[str] = None,
+    cache_dir: Optional[str] = None
+) -> Tuple[List[int], List[float], List[float], float, float, str, bool]:
+    """
+    Safe SyncNet inference with automatic audio extraction if needed.
+    
+    This function provides a wrapper around SyncNetPipeline.inference() that:
+    1. Handles audio_path=None by automatically extracting audio from video
+    2. Provides comprehensive error handling and logging
+    3. Ensures proper cleanup of temporary files
+    
+    Args:
+        pipeline: Initialized SyncNetPipeline instance
+        video_path: Path to input video file
+        audio_path: Path to audio file (None for auto-extraction from video)
+        cache_dir: Directory for temporary files (None for auto-cleanup)
+        
+    Returns:
+        Tuple of (offsets, confidences, distances, max_conf, min_dist, s3fd_json, has_face)
+        
+    Raises:
+        RuntimeError: If processing fails at any stage
+        FileNotFoundError: If video file doesn't exist
+        ValueError: If video format is not supported
+    """
+    if not os.path.exists(video_path):
+        raise FileNotFoundError(f"Video file not found: {video_path}")
+    
+    if audio_path and not os.path.exists(audio_path):
+        raise FileNotFoundError(f"Audio file not found: {audio_path}")
+    
+    logging.info(f"Starting safe SyncNet inference for video: {video_path}")
+    
+    try:
+        # Use the updated inference method that handles audio_path=None
+        results = pipeline.inference(
+            video_path=video_path,
+            audio_path=audio_path,
+            cache_dir=cache_dir
+        )
+        
+        offsets, confs, dists, max_conf, min_dist, s3fd_json, has_face = results
+        
+        # Validate results
+        if not has_face:
+            logging.warning("No faces detected in the video")
+        elif not offsets:
+            logging.warning("No valid face tracks found")
+        else:
+            logging.info(f"Successfully processed {len(offsets)} face tracks")
+            logging.info(f"LSE-C (max confidence): {max_conf:.3f}")
+            logging.info(f"LSE-D (min distance): {min_dist:.3f}")
+        
+        return results
+        
+    except Exception as e:
+        logging.error(f"SyncNet inference failed: {str(e)}")
+        raise RuntimeError(f"SyncNet processing failed: {str(e)}") from e
+
+
+def extract_audio_from_video(video_path: str, audio_path: str) -> None:
+    """
+    Extract audio from video file using ffmpeg.
+    
+    Args:
+        video_path: Path to input video file
+        audio_path: Path for output audio file
+        
+    Raises:
+        RuntimeError: If audio extraction fails
+        FileNotFoundError: If video file doesn't exist
+    """
+    if not os.path.exists(video_path):
+        raise FileNotFoundError(f"Video file not found: {video_path}")
+    
+    # Create a temporary pipeline instance just for audio extraction
+    pipeline = SyncNetPipeline()
+    
+    try:
+        pipeline._extract_audio_from_video(video_path, audio_path)
+        logging.info(f"Successfully extracted audio from {video_path} to {audio_path}")
+    except Exception as e:
+        logging.error(f"Audio extraction failed: {str(e)}")
+        raise RuntimeError(f"Audio extraction failed: {str(e)}") from e
+
+
+def calculate_lse_metrics(
+    pipeline: SyncNetPipeline,
+    video_path: str,
+    audio_path: Optional[str] = None,
+    cache_dir: Optional[str] = None
+) -> Tuple[float, float, str]:
+    """
+    Calculate LSE-C and LSE-D metrics for a video.
+    
+    Args:
+        pipeline: Initialized SyncNetPipeline instance
+        video_path: Path to input video file
+        audio_path: Path to audio file (None for auto-extraction)
+        cache_dir: Directory for temporary files
+        
+    Returns:
+        Tuple of (lse_c, lse_d, quality_assessment)
+        
+    Raises:
+        RuntimeError: If metric calculation fails
+    """
+    try:
+        results = safe_syncnet_inference(pipeline, video_path, audio_path, cache_dir)
+        offsets, confs, dists, max_conf, min_dist, s3fd_json, has_face = results
+        
+        if not has_face:
+            return 0.0, float('inf'), "NO_FACE"
+        
+        if not confs or not dists:
+            return 0.0, float('inf'), "NO_TRACKS"
+        
+        # LSE-C is the maximum confidence across all tracks
+        lse_c = max_conf
+        
+        # LSE-D is the minimum distance across all tracks  
+        lse_d = min_dist
+        
+        # Quality assessment based on typical thresholds
+        if lse_c > 3.5 and lse_d < 7.0:
+            quality = "GOOD"
+        elif lse_c > 2.0 and lse_d < 10.0:
+            quality = "FAIR"
+        else:
+            quality = "POOR"
+        
+        logging.info(f"LSE Metrics - C: {lse_c:.3f}, D: {lse_d:.3f}, Quality: {quality}")
+        
+        return lse_c, lse_d, quality
+        
+    except Exception as e:
+        logging.error(f"LSE metric calculation failed: {str(e)}")
+        raise RuntimeError(f"LSE metric calculation failed: {str(e)}") from e

syncnet_python/syncnet_pipeline.py

@@ -1,3 +1,14 @@
+# Modified from MoChaBench eval-lipsync/script/syncnet_pipeline.py
+# (https://github.com/congwei1230/MoChaBench, Apache-2.0) by nawta:
+# package-relative imports with a fallback for direct script execution;
+# ffmpeg_bin defaults to "ffmpeg" when None; the two shell
+# subprocess.call ffmpeg commands in _crop became subprocess.run argument
+# lists that raise RuntimeError on failure; added _extract_audio_from_video
+# so inference() accepts audio_path=None; the first three ffmpeg-python
+# calls in inference() raise RuntimeError on ffmpeg.Error; inference()
+# raises RuntimeError when no frames are extracted.
+# See NOTICE for licensing.
+
 import json
 import logging
 import os
@@ -215,11 +226,34 @@ class SyncNetPipeline:
         os.remove(tmp_avi)
         return final_avi
 
+    # ---------------------------- audio extraction helper ----------------- #
+    def _extract_audio_from_video(self, video_path: str, output_path: str) -> None:
+        """Extract audio from video file using ffmpeg."""
+        cfg = self.cfg
+        ffmpeg_bin = cfg.ffmpeg_bin if cfg.ffmpeg_bin is not None else "ffmpeg"
+        
+        cmd = [
+            ffmpeg_bin, "-y", "-i", str(video_path),
+            "-ac", "1", "-ar", str(cfg.audio_sample_rate),
+            "-acodec", "pcm_s16le", "-f", "wav",
+            str(output_path)
+        ]
+        
+        try:
+            result = subprocess.run(cmd, capture_output=True, text=True, check=True)
+            logging.info(f"Successfully extracted audio from {video_path} to {output_path}")
+        except subprocess.CalledProcessError as e:
+            logging.error(f"FFmpeg audio extraction failed: {e.stderr}")
+            raise RuntimeError(f"FFmpeg audio extraction failed: {e.stderr}")
+        except FileNotFoundError:
+            logging.error(f"FFmpeg not found at: {ffmpeg_bin}")
+            raise RuntimeError(f"FFmpeg not found. Please ensure ffmpeg is installed and in PATH.")
+
     # ---------------------------- inference -------------------------------- #
     def inference(
         self,
-        video_path: str,  # We do not extract audio from video_path!
-        audio_path: str,
+        video_path: str,
+        audio_path: Optional[str] = None,  # Now supports None for auto-extraction
         *,
         cache_dir: Optional[str] = None,
     ) -> Tuple[List[int], List[float], List[float], float, float, str, bool]:
@@ -229,34 +263,59 @@ class SyncNetPipeline:
             work.mkdir(parents=True, exist_ok=True)
 
         try:
+            # Handle audio_path=None case - extract audio from video
+            if audio_path is None:
+                logging.info("audio_path is None, extracting audio from video")
+                extracted_audio_path = work / "extracted_audio.wav"
+                self._extract_audio_from_video(video_path, str(extracted_audio_path))
+                actual_audio_path = str(extracted_audio_path)
+            else:
+                actual_audio_path = audio_path
+                logging.info(f"Using provided audio path: {actual_audio_path}")
+
             # 1) Convert video to constant-fps AVI
             avi = work / "video.avi"
-            (
-                ffmpeg.input(video_path)
-                .output(str(avi), **{"q:v": 2}, r=cfg.frame_rate, **{"async": 1})
-                .overwrite_output()
-                .run()
-            )
+            try:
+                (
+                    ffmpeg.input(video_path)
+                    .output(str(avi), **{"q:v": 2}, r=cfg.frame_rate, **{"async": 1})
+                    .overwrite_output()
+                    .run()
+                )
+            except ffmpeg.Error as e:
+                logging.error(f"FFmpeg video conversion failed: {e}")
+                raise RuntimeError(f"FFmpeg video conversion failed: {e}")
 
             # 2) Extract frames
             frames_dir = work / "frames"
             frames_dir.mkdir(exist_ok=True)
-            (
-                ffmpeg.input(str(avi))
-                .output(str(frames_dir / "%06d.jpg"), **{"q:v": 2}, f="image2", threads=1)
-                .overwrite_output()
-                .run()
-            )
+            try:
+                (
+                    ffmpeg.input(str(avi))
+                    .output(str(frames_dir / "%06d.jpg"), **{"q:v": 2}, f="image2", threads=1)
+                    .overwrite_output()
+                    .run()
+                )
+            except ffmpeg.Error as e:
+                logging.error(f"FFmpeg frame extraction failed: {e}")
+                raise RuntimeError(f"FFmpeg frame extraction failed: {e}")
+            
             frames = sorted(glob(str(frames_dir / "*.jpg")))
+            if not frames:
+                raise RuntimeError("No frames were extracted from the video")
 
             # 3) Resample speech
             audio_wav = work / "speech.wav"
-            (
-                ffmpeg.input(audio_path)
-                .output(str(audio_wav), ac=1, ar=cfg.audio_sample_rate, format="wav")
-                .overwrite_output()
-                .run()
-            )
+            try:
+                (
+                    ffmpeg.input(actual_audio_path)
+                    .output(str(audio_wav), ac=1, ar=cfg.audio_sample_rate, format="wav")
+                    .overwrite_output()
+                    .run()
+                )
+            except ffmpeg.Error as e:
+                logging.error(f"FFmpeg audio resampling failed: {e}")
+                raise RuntimeError(f"FFmpeg audio resampling failed: {e}")
 
             # 4) Face detection
             detections = []

syncnet_python/test_error_handling.py

@@ -0,0 +1,141 @@
+#!/usr/bin/env python3
+"""Test script for comprehensive error handling in SyncNet v0.2.2."""
+
+import logging
+import tempfile
+from pathlib import Path
+import sys
+import os
+sys.path.insert(0, os.path.dirname(__file__))
+
+from syncnet_pipeline import SyncNetPipeline
+
+# Import safe utils with fallback
+try:
+    from safe_syncnet_utils import safe_syncnet_inference, calculate_lse_metrics
+except ImportError:
+    # Fallback implementation for testing
+    def safe_syncnet_inference(pipeline, video_path, audio_path=None, cache_dir=None):
+        return pipeline.inference(video_path=video_path, audio_path=audio_path, cache_dir=cache_dir)
+    
+    def calculate_lse_metrics(pipeline, video_path, audio_path=None, cache_dir=None):
+        results = pipeline.inference(video_path=video_path, audio_path=audio_path, cache_dir=cache_dir)
+        offsets, confs, dists, max_conf, min_dist, _, has_face = results
+        
+        if not has_face or not confs:
+            return 0.0, float('inf'), "NO_FACE"
+        
+        quality = "GOOD" if max_conf > 3.5 and min_dist < 7.0 else "FAIR" if max_conf > 2.0 else "POOR"
+        return max_conf, min_dist, quality
+
+# Configure logging
+logging.basicConfig(
+    level=logging.INFO,
+    format="%(asctime)s [%(levelname)s] %(message)s"
+)
+
+def test_audio_none_handling():
+    """Test automatic audio extraction when audio_path=None."""
+    print("\n=== Testing audio_path=None handling ===")
+    
+    # Initialize pipeline
+    pipeline = SyncNetPipeline(
+        {'s3fd_weights': '../weights/sfd_face.pth', 'syncnet_weights': '../weights/syncnet_v2.model'},
+        device='cuda'
+    )
+    
+    # Test with audio_path=None (should automatically extract audio)
+    video_path = '../example/pair_0000_lipsynced.mp4'
+    
+    try:
+        print(f"Testing inference with audio_path=None for: {video_path}")
+        results = pipeline.inference(
+            video_path=video_path,
+            audio_path=None,  # This should now work!
+            cache_dir='../example/cache_test_none'
+        )
+        
+        offsets, confs, dists, max_conf, min_dist, _, has_face = results
+        
+        if has_face:
+            print(f"✅ SUCCESS: Auto audio extraction worked!")
+            print(f"   LSE-C: {max_conf:.3f}")
+            print(f"   LSE-D: {min_dist:.3f}")
+        else:
+            print("⚠️  No face detected, but no error occurred")
+            
+    except Exception as e:
+        print(f"❌ FAILED: {str(e)}")
+
+def test_safe_wrapper_functions():
+    """Test the safe wrapper functions."""
+    print("\n=== Testing Safe Wrapper Functions ===")
+    
+    pipeline = SyncNetPipeline(
+        {'s3fd_weights': '../weights/sfd_face.pth', 'syncnet_weights': '../weights/syncnet_v2.model'},
+        device='cuda'
+    )
+    
+    video_path = '../example/pair_0000_lipsynced.mp4'
+    
+    try:
+        print("Testing safe_syncnet_inference...")
+        results = safe_syncnet_inference(
+            pipeline, 
+            video_path, 
+            audio_path=None,
+            cache_dir='../example/cache_test_safe'
+        )
+        print("✅ safe_syncnet_inference: SUCCESS")
+        
+        print("Testing calculate_lse_metrics...")
+        lse_c, lse_d, quality = calculate_lse_metrics(
+            pipeline,
+            video_path,
+            audio_path=None,
+            cache_dir='../example/cache_test_metrics'
+        )
+        print(f"✅ calculate_lse_metrics: SUCCESS")
+        print(f"   LSE-C: {lse_c:.3f}")
+        print(f"   LSE-D: {lse_d:.3f}")
+        print(f"   Quality: {quality}")
+        
+    except Exception as e:
+        print(f"❌ Safe wrapper test FAILED: {str(e)}")
+
+def test_error_cases():
+    """Test various error cases."""
+    print("\n=== Testing Error Cases ===")
+    
+    pipeline = SyncNetPipeline(
+        {'s3fd_weights': '../weights/sfd_face.pth', 'syncnet_weights': '../weights/syncnet_v2.model'},
+        device='cuda'
+    )
+    
+    # Test with non-existent file
+    try:
+        results = safe_syncnet_inference(
+            pipeline,
+            "non_existent_video.mp4",
+            audio_path=None
+        )
+        print("❌ Should have failed for non-existent file")
+    except FileNotFoundError:
+        print("✅ Correctly caught FileNotFoundError for non-existent video")
+    except Exception as e:
+        print(f"⚠️  Unexpected error: {str(e)}")
+
+def main():
+    """Run all tests."""
+    print("SyncNet v0.2.2 Error Handling Test Suite")
+    print("=" * 50)
+    
+    test_audio_none_handling()
+    test_safe_wrapper_functions() 
+    test_error_cases()
+    
+    print("\n" + "=" * 50)
+    print("Test suite completed!")
+
+if __name__ == "__main__":
+    main()

syncnet_python-0.2.3.dist-info/METADATA

@@ -0,0 +1,204 @@
+Metadata-Version: 2.4
+Name: syncnet-python
+Version: 0.2.3
+Summary: SyncNet: Audio-visual synchronization detection using deep learning. Updated version of https://github.com/joonson/syncnet_python for modern Python versions.
+Author: SyncNet Python Contributors
+Maintainer: SyncNet Python Contributors
+License-Expression: MIT AND Apache-2.0
+Project-URL: Homepage, https://github.com/nawta/SyncNet_py309_313
+Project-URL: Repository, https://github.com/nawta/SyncNet_py309_313
+Project-URL: Issues, https://github.com/nawta/SyncNet_py309_313/issues
+Keywords: audio-visual,synchronization,deep-learning,pytorch,lip-sync,video-processing,computer-vision
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Science/Research
+Classifier: Intended Audience :: Developers
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.9
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Topic :: Multimedia :: Video
+Classifier: Topic :: Scientific/Engineering :: Image Processing
+Requires-Python: >=3.9
+Description-Content-Type: text/markdown
+License-File: LICENSE
+License-File: LICENSE-APACHE
+License-File: NOTICE
+Requires-Dist: torch>=2.0.0
+Requires-Dist: torchvision>=0.15.0
+Requires-Dist: numpy>=1.24.0
+Requires-Dist: scipy>=1.10.0
+Requires-Dist: pandas>=2.0.0
+Requires-Dist: scenedetect<0.7,>=0.6.0
+Requires-Dist: opencv-contrib-python>=4.8.0
+Requires-Dist: python-speech-features>=0.6
+Requires-Dist: ffmpeg-python>=0.2.0
+Provides-Extra: dev
+Requires-Dist: pytest>=7.4.0; extra == "dev"
+Requires-Dist: pytest-asyncio>=0.21.0; extra == "dev"
+Requires-Dist: black>=23.0.0; extra == "dev"
+Requires-Dist: ruff>=0.1.0; extra == "dev"
+Requires-Dist: mypy>=1.7.0; extra == "dev"
+Requires-Dist: types-opencv-python; extra == "dev"
+Requires-Dist: types-scipy; extra == "dev"
+Dynamic: license-file
+
+# syncnet-python
+
+[![PyPI version](https://badge.fury.io/py/syncnet-python.svg)](https://badge.fury.io/py/syncnet-python)
+[![Python](https://img.shields.io/pypi/pyversions/syncnet-python.svg)](https://pypi.org/project/syncnet-python/)
+[![License: MIT AND Apache-2.0](https://img.shields.io/badge/License-MIT%20AND%20Apache--2.0-yellow.svg)](https://github.com/nawta/SyncNet_py309_313/blob/main/NOTICE)
+
+A pip-installable SyncNet for Python 3.9 to 3.13 and PyTorch 2. It computes the confidence and minimum distance scores that talking-head and lip-sync papers report as LSE-C and LSE-D.
+
+SyncNet (Chung and Zisserman, 2016) is a network that measures how well mouth movements in a video match the speech audio. This package wraps the original SyncNet model and its S3FD face detector in one Python class and one command. Given a video, it finds and tracks faces, crops each face track, and returns three numbers per track: the audio-video offset in frames, the SyncNet confidence (LSE-C, higher means better sync), and the minimum audio-video feature distance (LSE-D, lower means better sync).
+
+## Quickstart
+
+### 1. Install
+
+```bash
+pip install syncnet-python
+```
+
+You also need `ffmpeg` on your `PATH` (`brew install ffmpeg` or `sudo apt-get install ffmpeg`).
+
+Use version 0.2.3 or later. Versions 0.2.2 and earlier accept scenedetect 0.7, which removed the `scenedetect.video_manager` module the pipeline imports, and then `from syncnet_python import SyncNetPipeline` silently gives `None`. Version 0.2.3 requires `scenedetect>=0.6,<0.7` and raises an `ImportError` with the underlying message if an import fails. If you must stay on 0.2.2, install it with `pip install syncnet-python==0.2.2 "scenedetect<0.7"`.
+
+The package uses `opencv-contrib-python` for `cv2`. Since 0.2.3 it is the only OpenCV package the install pulls in. Versions up to 0.2.2 also installed `opencv-python` through `scenedetect[opencv]`; if you upgrade an old environment and `cv2` fails to import, run `pip uninstall opencv-python opencv-contrib-python` and then `pip install opencv-contrib-python`.
+
+### 2. Download the weights
+
+The package does not include the model weights. Both files come from the Oxford VGG SyncNet page, the same URLs that the original repository's `download_model.sh` uses:
+
+```bash
+mkdir -p weights
+wget https://www.robots.ox.ac.uk/~vgg/software/lipsync/data/syncnet_v2.model -O weights/syncnet_v2.model
+wget https://www.robots.ox.ac.uk/~vgg/software/lipsync/data/sfd_face.pth -O weights/sfd_face.pth
+```
+
+The same two files are also in the [`weights/`](https://github.com/nawta/SyncNet_py309_313/tree/main/weights) folder of this repository. They are byte-identical to the Oxford files (SHA-256 `961e8696…5442` for `syncnet_v2.model` and `d54a87c2…c491` for `sfd_face.pth`).
+
+### 3. Python API
+
+```python
+from syncnet_python import SyncNetPipeline
+
+pipe = SyncNetPipeline(
+    {"s3fd_weights": "weights/sfd_face.pth", "syncnet_weights": "weights/syncnet_v2.model"},
+    device="cpu",  # or "cuda"
+)
+offsets, confs, dists, best_conf, min_dist, s3fd_json, has_face = pipe.inference(
+    video_path="clip.mp4",
+    audio_path=None,  # None uses the video's own audio track; or pass a .wav path
+)
+print("AV offset (frames):", offsets[0])
+print("LSE-C (confidence):", round(best_conf, 3))
+print("LSE-D (min distance):", round(min_dist, 3))
+```
+
+`offsets`, `confs` and `dists` are lists with one entry per face track. `best_conf` is the largest confidence over all tracks and `min_dist` is the smallest distance over all tracks. For a video with one face these equal `confs[0]` and `dists[0]`. With several faces the two values can come from different tracks, so use the per-track lists if you need scores for one speaker.
+
+`syncnet_python.calculate_lse_metrics(pipe, video_path)` returns `(lse_c, lse_d, quality_label)` from the same values.
+
+### 4. Command line
+
+```bash
+syncnet-python clip.mp4 --device cpu -o results.json
+```
+
+The command reads the weights from `weights/` in the current directory by default; change this with `--s3fd-weights` and `--syncnet-weights`. It prints the offset and confidence for each video and writes offset, confidence and minimum distance to the JSON file. The default device is `cuda`, so pass `--device cpu` on a machine without an NVIDIA GPU.
+
+### Tested setup
+
+The examples above were run on 2026-10-06 on an Apple Silicon Mac (CPU), using [`example/video.avi`](https://github.com/nawta/SyncNet_py309_313/blob/main/example/video.avi) converted to MP4. Each run used a fresh environment. The OpenCV column gives the version of `opencv-contrib-python`. In the 0.2.3 runs it was the only OpenCV package installed; in the 0.2.2 run `opencv-python` was also installed at the same version.
+
+| Package | Python | PyTorch | NumPy | OpenCV | scenedetect | Offset | LSE-C | LSE-D |
+|---|---|---|---|---|---|---|---|---|
+| 0.2.3 | 3.13 (arm64) | 2.14.1 | 2.5.3 | 5.0.0 | 0.6.7.1 | 1 | 4.529 | 9.237 |
+| 0.2.3 | 3.10 (arm64) | 2.14.1 | 2.2.6 | 5.0.0 | 0.6.7.1 | 1 | 4.529 | 9.237 |
+| 0.2.2 | 3.9 (x86_64 under Rosetta 2) | 2.2.2 | 1.26.4 | 4.11.0 | 0.6.7.1 | 1 | 4.524 | 9.291 |
+
+The 0.2.3 rows used the wheel built from this repository with a plain install. The Python API, `calculate_lse_metrics` and the CLI gave the same values. The Python 3.9 interpreter was an x86_64 build running under Rosetta 2. PyTorch 2.2.2 is the newest release with macOS x86_64 wheels, and it does not work with NumPy 2, so that run needed `numpy<2`, `opencv-python<4.12` and `opencv-contrib-python<4.12` installed by hand. The full run on the 5.3-second clip took about 10 seconds on CPU.
+
+## Changes in 0.2.3
+
+- Requires `scenedetect>=0.6,<0.7`, so a plain `pip install syncnet-python` works again.
+- A failed import inside the package raises `ImportError` with the original message. Earlier versions set `SyncNetPipeline` and the other exports to `None`.
+- The repository now holds the 0.2.2 code that was published on PyPI but not committed (`calculate_lse_metrics`, `audio_path=None`, `ffmpeg` error handling).
+- Package metadata declares the license as `MIT AND Apache-2.0` and ships `LICENSE`, `LICENSE-APACHE` and `NOTICE`.
+
+See [CHANGELOG.md](https://github.com/nawta/SyncNet_py309_313/blob/main/CHANGELOG.md) for earlier versions.
+
+## Comparison with joonson/syncnet_python
+
+The original repository, [joonson/syncnet_python](https://github.com/joonson/syncnet_python), was updated by its author on 2026-04-17 (PR #78). This table compares that version with this package.
+
+| | joonson/syncnet_python (2026-04) | syncnet-python 0.2.3 |
+|---|---|---|
+| Install | clone, then `conda env create -f environment.yml`; no `setup.py` or `pyproject.toml` | `pip install syncnet-python` |
+| Python | 3.10 (pinned in `environment.yml`) | 3.9 to 3.13 (3.9, 3.10 and 3.13 tested above) |
+| PyTorch | 2.5.1 (pinned) | `torch>=2.0.0` |
+| scenedetect | 0.6.7.1 (pinned) | `>=0.6,<0.7` |
+| Python API | `SyncNetInstance.evaluate()` and `extract_feature()` score a pre-cropped face clip; face detection, tracking and cropping run only through `run_pipeline.py` | `SyncNetPipeline(...).inference(video_path, audio_path)` runs face detection, tracking, cropping and scoring in one call |
+| Command line | `run_pipeline.py`, `run_syncnet.py`, `run_visualise.py` run in sequence, plus `demo_syncnet.py` for pre-cropped clips | one `syncnet-python` command that runs detection, tracking, cropping and scoring |
+| Output | offset, minimum distance and confidence written to the log; per-frame distances saved as `activesd.pckl` under `--data_dir` | values returned to Python, or written to JSON by the CLI |
+| Weights | downloaded by `download_model.sh` | downloaded separately (see above) |
+| Visualisation of the result | `run_visualise.py` | not included |
+
+## Errors this fixes
+
+Before the April 2026 update, the original repository pinned `scenedetect==0.5.1` and used `np.int`. Users hit these two errors in `run_pipeline.py`:
+
+```
+TypeError: 'tuple' object does not support item assignment
+```
+
+This comes from scenedetect 0.5.1's `ContentDetector` running with newer OpenCV (upstream issues [#55](https://github.com/joonson/syncnet_python/issues/55) and [#69](https://github.com/joonson/syncnet_python/issues/69)). This package uses scenedetect 0.6.
+
+```
+AttributeError: module 'numpy' has no attribute 'int'.
+```
+
+This comes from `.astype(np.int)` in `detectors/s3fd/box_utils.py`. NumPy 1.24 removed `np.int`. This package uses `.astype(int)`.
+
+## Repository layout
+
+The PyPI package contains only the `syncnet_python/` folder. Most of its files (`syncnet_pipeline.py`, `SyncNetInstance.py`, `SyncNetModel.py`, `detectors/` and the scripts named `run_syncnet_pipeline_on_*.py`) come from the SyncNet evaluation code in [MoChaBench](https://github.com/congwei1230/MoChaBench) (`eval-lipsync/script/`), which builds on the original repository. This package adds error handling around the `ffmpeg` calls in `syncnet_pipeline.py`, plus `cli.py`.
+
+The `syncnet/` folder holds a separate refactor with configuration files, logging and batch helpers. It is in this repository only and pip does not install it. `scripts/` has example scripts, and `example/` has a short test video with its audio.
+
+## Credits
+
+The SyncNet model, its pretrained weights and the original code are by Joon Son Chung and Andrew Zisserman ([joonson/syncnet_python](https://github.com/joonson/syncnet_python), [project page](https://www.robots.ox.ac.uk/~vgg/software/lipsync/)). The S3FD face detector weights (`sfd_face.pth`) are downloaded from the same page. The model, detector and pipeline files in `syncnet_python/` come from [MoChaBench](https://github.com/congwei1230/MoChaBench); [NOTICE](https://github.com/nawta/SyncNet_py309_313/blob/main/NOTICE) lists them.
+
+## Citation
+
+If you use this code in your research, please cite the original paper:
+
+```bibtex
+@InProceedings{Chung16a,
+  author       = "Chung, J.~S. and Zisserman, A.",
+  title        = "Out of time: automated lip sync in the wild",
+  booktitle    = "Workshop on Multi-view Lip-reading, ACCV",
+  year         = "2016",
+}
+```
+
+## License
+
+This repository uses two licenses. [NOTICE](https://github.com/nawta/SyncNet_py309_313/blob/main/NOTICE) lists which file falls under which.
+
+- The original SyncNet code by Joon Son Chung and the code written for this repository are under the MIT License ([LICENSE](https://github.com/nawta/SyncNet_py309_313/blob/main/LICENSE)).
+- The files taken from MoChaBench are under the Apache License 2.0 ([LICENSE-APACHE](https://github.com/nawta/SyncNet_py309_313/blob/main/LICENSE-APACHE)).
+
+The model weights have their own terms. The Oxford VGG SyncNet page says: "The model can be used for research purposes under Creative Commons Attribution License." The MIT and Apache licenses above do not cover the weights.
+
+## Links
+
+- Source and issues: https://github.com/nawta/SyncNet_py309_313
+- PyPI: https://pypi.org/project/syncnet-python/
+- Original SyncNet: https://github.com/joonson/syncnet_python

{syncnet_python-0.2.1.dist-info → syncnet_python-0.2.3.dist-info}/RECORD

@@ -1,19 +1,23 @@
 syncnet_python/SyncNetInstance.py,sha256=V-RtWL4nN2CW5n56DkE0d6hcnet4hAs_OVMaMS1ih20,6227
 syncnet_python/SyncNetModel.py,sha256=6qk27paoyV39MTsVyu6K2sbbM-1SLU_2N1zbqm1eeYs,3575
-syncnet_python/__init__.py,sha256=BAgXJdworzKFgKVrUD7cncAhhKlrqF5nMSnOcpTUvjM,702
+syncnet_python/__init__.py,sha256=4OhqOByDwnM2ZhqaE2y4L02kPr5VRhhCEVSLck9QxSs,1241
 syncnet_python/cli.py,sha256=YSQVVDChLkQ96gFRFqD7laQXqbrLC2b8SKaF-vc8FuA,3256
 syncnet_python/run_syncnet_pipeline_on_1example.py,sha256=7I8_jEgIFw-RUqCHU29LrIWIRP77lTUiH6MePOxsuWQ,1008
 syncnet_python/run_syncnet_pipeline_on_mocha_generation_on_mocha_bench.py,sha256=HJHGLgKpf21bG27_x1QD3MAhqtHey26Ujxb8s2aK-tA,5928
 syncnet_python/run_syncnet_pipeline_on_your_own_model_results.py,sha256=m-kz_DHhka_DywpBRWdxk-S0hrwqvCR16RiVQ7HUqDk,6003
-syncnet_python/syncnet_pipeline.py,sha256=FJwhJ9_brdlY73gSkVbKkdZm197pNkcc3FAKyr-ubpY,13263
+syncnet_python/safe_syncnet_utils.py,sha256=viucLggg13AW5xAjns_s-jlqv4rzvuRED0Ydo99_RbY,5381
+syncnet_python/syncnet_pipeline.py,sha256=G9msGO9JDzfCkLD4sBQ8SOlk9QfCEPqjb9mbVAl93SA,16410
+syncnet_python/test_error_handling.py,sha256=i5ib1qEjbMUQg913r9plAdXYciQx2jmUjzuoXPvc5f4,4658
 syncnet_python/test_lse_metrics.py,sha256=H_4U2LN-Wussx_KABX5y5EO3QGnaVhKNMuTjLH0zNrE,2884
 syncnet_python/detectors/__init__.py,sha256=WLone-DTbvQUFleY93pmq7xomPDVjda9ATpZYcS6_sA,23
 syncnet_python/detectors/s3fd/__init__.py,sha256=MIJfIEsKFTGh8brmD3fBQg8JZAp57NtaTkY6_I5QZP0,2322
 syncnet_python/detectors/s3fd/box_utils.py,sha256=CWn46LMJKO1cdenI_IWxBbDriO2ueWTCl14Xp28fr-U,7235
 syncnet_python/detectors/s3fd/nets.py,sha256=BYPJq9UJ5bq5Q5RgP1j-NsFzO1gP3N0PYE1X19-4PZw,5891
-syncnet_python-0.2.1.dist-info/licenses/LICENSE,sha256=qwJOQjZqnGgzcgwg3VUkT2pGKwLyZr5nUOm0Gv7Zfog,1088
-syncnet_python-0.2.1.dist-info/METADATA,sha256=3l1AsKqZM6hI1bQKaj6zeThFgSUi-JPgsF1iF6Ct3PY,7747
-syncnet_python-0.2.1.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-syncnet_python-0.2.1.dist-info/entry_points.txt,sha256=zPJrIl_ekAa31dHZNCoTHiII2k_m5GyMD5BEkMR6Ytk,59
-syncnet_python-0.2.1.dist-info/top_level.txt,sha256=OkvKpxwq9NzQSjjdD571umGrI3meW9kVleOV2N5Ua04,15
-syncnet_python-0.2.1.dist-info/RECORD,,
+syncnet_python-0.2.3.dist-info/licenses/LICENSE,sha256=JMCE5BdW3pk3NVn8ksaNiLxEABgPjlLkW-3ERoLPhug,1170
+syncnet_python-0.2.3.dist-info/licenses/LICENSE-APACHE,sha256=z8d0m5b2O9McPEK1xHG_dWgUBT6EfBDz6wA0F7xSPTA,11358
+syncnet_python-0.2.3.dist-info/licenses/NOTICE,sha256=rtNvKHi6F6xy8MOG15aetzhkOt8DqdPcswsqCb-msLs,2033
+syncnet_python-0.2.3.dist-info/METADATA,sha256=wM7GvQfaJxBJ4JUtl_m8vyX5s2z2tEysI7UCCQq5QMk,13109
+syncnet_python-0.2.3.dist-info/WHEEL,sha256=YVMoNqKzERt-wjUZwJ33xBGAwnFl-4cqbYkTtWa4itE,91
+syncnet_python-0.2.3.dist-info/entry_points.txt,sha256=zPJrIl_ekAa31dHZNCoTHiII2k_m5GyMD5BEkMR6Ytk,59
+syncnet_python-0.2.3.dist-info/top_level.txt,sha256=OkvKpxwq9NzQSjjdD571umGrI3meW9kVleOV2N5Ua04,15
+syncnet_python-0.2.3.dist-info/RECORD,,

{syncnet_python-0.2.1.dist-info → syncnet_python-0.2.3.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (80.9.0)
+Generator: setuptools (84.0.0)
 Root-Is-Purelib: true
 Tag: py3-none-any
 

{syncnet_python-0.2.1.dist-info → syncnet_python-0.2.3.dist-info}/licenses/LICENSE

@@ -1,6 +1,7 @@
 MIT License
 
-Copyright (c) 2024 SyncNet Python 3.13 Contributors
+Copyright (c) 2016-present Joon Son Chung (original SyncNet code)
+Copyright (c) 2024 SyncNet Python 3.13 Contributors (modifications)
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal

syncnet_python-0.2.3.dist-info/licenses/LICENSE-APACHE

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

syncnet_python-0.2.3.dist-info/licenses/NOTICE

@@ -0,0 +1,43 @@
+syncnet-python
+https://github.com/nawta/SyncNet_py309_313
+
+This repository contains code under two licenses.
+
+1. MIT License (see LICENSE)
+   The original SyncNet code by Joon Son Chung
+   (https://github.com/joonson/syncnet_python) and the code written for
+   this repository: syncnet_python/__init__.py, syncnet_python/cli.py,
+   syncnet_python/safe_syncnet_utils.py,
+   syncnet_python/test_error_handling.py,
+   syncnet_python/test_lse_metrics.py, and the syncnet/, scripts/ and
+   config/ folders.
+
+2. Apache License 2.0 (see LICENSE-APACHE)
+   The following files come from MoChaBench
+   (https://github.com/congwei1230/MoChaBench, folder eval-lipsync/script/),
+   which is distributed under the Apache License 2.0. MoChaBench's LICENSE
+   file does not name a copyright holder; the files are credited here to
+   the MoChaBench authors. MoChaBench's versions are in turn based on the
+   original SyncNet code above. All files in this list except
+   syncnet_pipeline.py are unchanged copies of MoChaBench's files.
+
+     syncnet_python/syncnet_pipeline.py   (modified by nawta; the file
+                                           header lists the changes)
+     syncnet_python/SyncNetInstance.py
+     syncnet_python/SyncNetModel.py
+     syncnet_python/detectors/__init__.py
+     syncnet_python/detectors/s3fd/__init__.py
+     syncnet_python/detectors/s3fd/box_utils.py
+     syncnet_python/detectors/s3fd/nets.py
+     syncnet_python/run_syncnet_pipeline_on_1example.py
+     syncnet_python/run_syncnet_pipeline_on_mocha_generation_on_mocha_bench.py
+     syncnet_python/run_syncnet_pipeline_on_your_own_model_results.py
+
+Model weights
+
+   weights/syncnet_v2.model and weights/sfd_face.pth are copies of the files
+   published by the Visual Geometry Group, University of Oxford, at
+   https://www.robots.ox.ac.uk/~vgg/software/lipsync/. That page states:
+   "The model can be used for research purposes under Creative Commons
+   Attribution License." Neither the MIT License nor the Apache License 2.0
+   above applies to these files.

syncnet_python-0.2.1.dist-info/METADATA

@@ -1,217 +0,0 @@
-Metadata-Version: 2.4
-Name: syncnet-python
-Version: 0.2.1
-Summary: SyncNet: Audio-visual synchronization detection using deep learning. Updated version of https://github.com/joonson/syncnet_python for modern Python versions.
-Author: SyncNet Python Contributors
-Maintainer: SyncNet Python Contributors
-License: MIT
-Project-URL: Homepage, https://github.com/nawta/SyncNet_py313
-Project-URL: Bug Reports, https://github.com/nawta/SyncNet_py313/issues
-Project-URL: Source, https://github.com/nawta/SyncNet_py313
-Keywords: audio-visual,synchronization,deep-learning,pytorch,lip-sync,video-processing,computer-vision
-Classifier: Development Status :: 4 - Beta
-Classifier: Intended Audience :: Science/Research
-Classifier: Intended Audience :: Developers
-Classifier: License :: OSI Approved :: MIT License
-Classifier: Operating System :: OS Independent
-Classifier: Programming Language :: Python :: 3
-Classifier: Programming Language :: Python :: 3.9
-Classifier: Programming Language :: Python :: 3.10
-Classifier: Programming Language :: Python :: 3.11
-Classifier: Programming Language :: Python :: 3.12
-Classifier: Programming Language :: Python :: 3.13
-Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
-Classifier: Topic :: Multimedia :: Video
-Classifier: Topic :: Scientific/Engineering :: Image Processing
-Requires-Python: >=3.9
-Description-Content-Type: text/markdown
-License-File: LICENSE
-Requires-Dist: torch>=2.0.0
-Requires-Dist: torchvision>=0.15.0
-Requires-Dist: numpy>=1.24.0
-Requires-Dist: scipy>=1.10.0
-Requires-Dist: pandas>=2.0.0
-Requires-Dist: scenedetect[opencv]>=0.6.0
-Requires-Dist: opencv-contrib-python>=4.8.0
-Requires-Dist: python-speech-features>=0.6
-Requires-Dist: ffmpeg-python>=0.2.0
-Provides-Extra: dev
-Requires-Dist: pytest>=7.4.0; extra == "dev"
-Requires-Dist: pytest-asyncio>=0.21.0; extra == "dev"
-Requires-Dist: black>=23.0.0; extra == "dev"
-Requires-Dist: ruff>=0.1.0; extra == "dev"
-Requires-Dist: mypy>=1.7.0; extra == "dev"
-Requires-Dist: types-opencv-python; extra == "dev"
-Requires-Dist: types-scipy; extra == "dev"
-Dynamic: license-file
-
-# SyncNet Python
-
-[![PyPI version](https://badge.fury.io/py/syncnet-python.svg)](https://badge.fury.io/py/syncnet-python)
-[![Python](https://img.shields.io/pypi/pyversions/syncnet-python.svg)](https://pypi.org/project/syncnet-python/)
-[![Downloads](https://pepy.tech/badge/syncnet-python)](https://pepy.tech/project/syncnet-python)
-[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
-
-Audio-visual synchronization detection using deep learning with modern Python architecture.
-
-This is a **refactored and enhanced version** of the original [SyncNet implementation](https://github.com/joonson/syncnet_python) by Joon Son Chung, updated for Python 3.9+ with clean architecture, comprehensive error handling, and performance optimizations.
-
-## Overview
-
-SyncNet Python is a PyTorch implementation of the SyncNet model, which detects audio-visual synchronization in videos. It can identify lip-sync errors by analyzing the correspondence between mouth movements and spoken audio.
-
-## Features
-
-### Core Functionality
-- 🎥 **Audio-Visual Sync Detection**: Accurately detect synchronization between audio and video
-- 🔍 **Face Detection**: Automatic face detection and tracking using S3FD
-- 📊 **Detailed Analysis**: Per-crop offsets, confidence scores, and minimum distances
-- 🚀 **Batch Processing**: Process multiple videos efficiently
-- 🐍 **Python API**: Easy-to-use Python interface with proper error handling
-
-### Architecture Improvements
-- 🏗️ **Clean Architecture**: Abstract base classes and factory patterns
-- ⚡ **Performance Optimized**: Parallel processing and memory management
-- 🛡️ **Robust Error Handling**: Comprehensive exception hierarchy
-- ⚙️ **Configuration Management**: YAML/JSON configuration support
-- 📝 **Advanced Logging**: Structured logging with progress tracking
-- 🔄 **Backward Compatibility**: Maintains compatibility with original API
-
-## Installation
-
-```bash
-pip install syncnet-python
-```
-
-### Additional Requirements
-
-1. **FFmpeg**: Required for video processing
-   ```bash
-   # Ubuntu/Debian
-   sudo apt-get install ffmpeg
-   
-   # macOS
-   brew install ffmpeg
-   ```
-
-2. **Model Weights**: Download pre-trained weights
-   - Download `sfd_face.pth` and `syncnet_v2.model`
-   - Place them in a `weights/` directory
-
-## Quick Start
-
-```python
-from syncnet_python import SyncNetPipeline
-
-# Initialize pipeline
-pipeline = SyncNetPipeline(
-    s3fd_weights="weights/sfd_face.pth",
-    syncnet_weights="weights/syncnet_v2.model",
-    device="cuda"  # or "cpu"
-)
-
-# Process video
-results = pipeline.inference(
-    video_path="video.mp4",
-    audio_path=None  # Extract from video
-)
-
-# Extract results (returns tuple)
-offset_list, confidence_list, min_dist_list, best_confidence, best_min_dist, detections_json, success = results
-
-# Get best results
-offset = offset_list[0]  # AV offset in frames
-confidence = confidence_list[0]  # Confidence score
-min_distance = min_dist_list[0]  # Minimum distance
-
-print(f"AV Offset: {offset} frames")
-print(f"Confidence: {confidence:.3f}")
-print(f"Min Distance: {min_distance:.3f}")
-```
-
-### Detailed Analysis
-
-```python
-# For detailed per-crop analysis
-for i, (offset, conf, dist) in enumerate(zip(offset_list, confidence_list, min_dist_list)):
-    print(f"Crop {i+1}: offset={offset}, confidence={conf:.3f}, min_dist={dist:.3f}")
-
-# Parse face detections
-import json
-detections = json.loads(detections_json)
-print(f"Total frames with face detection: {len(detections)}")
-```
-
-## Command Line Usage
-
-```bash
-# Process single video
-syncnet-python video.mp4
-
-# Process multiple videos
-syncnet-python video1.mp4 video2.mp4 --output results.json
-
-# Use CPU instead of GPU
-syncnet-python video.mp4 --device cpu
-```
-
-## Performance
-
-Tested with example files:
-- **Processing Speed**: 191.4 fps
-- **Face Detection**: 100% success rate
-- **Accuracy**: Detects 1-frame offsets with high confidence (4.5+)
-- **Compute Time**: ~0.65 seconds for 134 frames
-
-## Architecture
-
-### Refactored Core Modules
-- `syncnet/core/` - Modern refactored implementation
-  - `base.py` - Abstract base classes and interfaces
-  - `models.py` - Enhanced SyncNet model with factory pattern
-  - `audio.py` - MFCC audio processing with streaming support
-  - `video.py` - Parallel video processing with OpenCV
-  - `sync_analyzer.py` - Optimized sync analysis with caching
-  - `config.py` - Configuration management system
-  - `exceptions.py` - Comprehensive error handling
-  - `logging.py` - Advanced logging with progress tracking
-  - `utils.py` - Memory management and utility functions
-
-### Legacy Compatibility
-- `syncnet_python/` - Maintains original API compatibility
-- Full backward compatibility with existing code
-
-## Requirements
-
-- Python 3.9+ (tested on 3.13)
-- PyTorch 2.0+
-- CUDA (optional but recommended)
-- FFmpeg
-- Additional dependencies: OpenCV, SciPy, NumPy, pandas
-
-## Credits
-
-This package is based on the original [SyncNet implementation](https://github.com/joonson/syncnet_python) by Joon Son Chung, enhanced with modern Python architecture and performance optimizations.
-
-## Citation
-
-If you use this code in your research, please cite the original paper:
-
-```bibtex
-@inproceedings{chung2016out,
-  title={Out of time: automated lip sync in the wild},
-  author={Chung, Joon Son and Zisserman, Andrew},
-  booktitle={Asian Conference on Computer Vision},
-  year={2016}
-}
-```
-
-## License
-
-MIT License - see LICENSE file for details.
-
-## Links
-
-- GitHub: https://github.com/yourusername/syncnet-python
-- Documentation: https://syncnet-python.readthedocs.io
-- Issues: https://github.com/yourusername/syncnet-python/issues