Open source video conferencing app powered by LiveKit. Built with Django and React.
π(summary) support media files with bad streams
Florent Chehab3b3f992834c8
diff --git a/CHANGELOG.md b/CHANGELOG.mdindex 2df979fe..78e2c3f7 100644--- a/CHANGELOG.md+++ b/CHANGELOG.md@@ -13,6 +13,7 @@ and this project adheres to- π(front) fix frontend build failure- π(makefile) fix args in make test- π©Ή(backend) fix case-insensitive email deduplication in merge command+- π(summary) support media files with bad streams #1478## [1.22.0] - 2026-07-03diff --git a/src/summary/summary/core/file_service.py b/src/summary/summary/core/file_service.pyindex 9088ccc9..1b4437f4 100644--- a/src/summary/summary/core/file_service.py+++ b/src/summary/summary/core/file_service.py@@ -142,6 +142,7 @@ class MediaInfo:has_video: boolaudio_duration_seconds: float | Noneaudio_codec_name: str | None+ has_bad_stream: bool = Falsedef get_media_info(local_path: Path) -> MediaInfo:@@ -174,8 +175,9 @@ def get_media_info(local_path: Path) -> MediaInfo:data = json.loads(result.stdout)streams = data.get("streams", [])- has_audio = any(el["codec_type"] == "audio" for el in streams)- has_video = any(el["codec_type"] == "video" for el in streams)+ has_audio = any(el.get("codec_type") == "audio" for el in streams)+ has_video = any(el.get("codec_type") == "video" for el in streams)+ has_bad_stream = any(el.get("codec_type", None) is None for el in streams)audio_codec_name = next((stream.get("codec_name")@@ -193,10 +195,11 @@ def get_media_info(local_path: Path) -> MediaInfo:has_video=has_video,audio_duration_seconds=audio_duration_seconds,audio_codec_name=audio_codec_name,+ has_bad_stream=has_bad_stream,)-def extract_audio_from_video(media_info: MediaInfo) -> Path:+def extract_audio_from_media(media_info: MediaInfo) -> Path:"""Extracts the audio track from a video file and saves it as a separate audio file.Based on the provided audio codec,@@ -451,7 +454,12 @@ class FileService:if media_info.has_video:logger.info("Video file detected, extracting audio...")- processed_path = extract_audio_from_video(media_info)+ processed_path = extract_audio_from_media(media_info)+ # Bad streams may cause transcription issues on WhisperX,+ # So we extract the audio properly+ elif media_info.has_bad_stream:+ logger.info("Bad stream detected, extracting audio...")+ processed_path = extract_audio_from_media(media_info)else:processed_path = downloaded_pathdiff --git a/src/summary/tests/unit/test_file_service.py b/src/summary/tests/unit/test_file_service.pyindex 8f9a92dd..f7e6fc6d 100644--- a/src/summary/tests/unit/test_file_service.py+++ b/src/summary/tests/unit/test_file_service.py@@ -1,12 +1,15 @@"""Unit tests for the file service."""+import jsonfrom pathlib import Path+from unittest.mock import Mockimport pytest+from summary.core import file_servicefrom summary.core.file_service import (MediaInfo,- extract_audio_from_video,+ extract_audio_from_media,get_media_info,)@@ -110,12 +113,40 @@ def test_media_info_invalid_file() -> None:)+def test_media_info_ignores_empty_stream_entry(monkeypatch: pytest.MonkeyPatch) -> None:+ """Test stream parsing when ffprobe returns an empty stream object."""+ ffprobe_payload = {+ "programs": [],+ "stream_groups": [],+ "streams": [+ {"codec_name": "vorbis", "codec_type": "audio"},+ {},+ ],+ }++ run_mock = Mock(+ return_value=Mock(stdout=json.dumps(ffprobe_payload), stderr="", returncode=0)+ )+ monkeypatch.setattr(file_service.subprocess, "run", run_mock)+ monkeypatch.setattr(+ file_service, "get_media_duration_seconds", Mock(return_value=2.5)+ )++ media_info = get_media_info(BASE_PATH / "audio-sample-android-firefox.ogg")++ assert media_info.has_audio is True+ assert media_info.has_video is False+ assert media_info.has_bad_stream is True+ assert media_info.audio_codec_name == "vorbis"+ assert media_info.audio_duration_seconds == 2.5++def test_extract_audio_from_video():"""Test that extract_audio_from_video can extract audio from a video file."""path = None# A bit of cleanup logic since this is not a generatortry:- path = extract_audio_from_video(MEDIA_INFO_SAMPLE_VISIO)+ path = extract_audio_from_media(MEDIA_INFO_SAMPLE_VISIO)assert path.name.endswith(".m4a")except Exception as e:pytest.fail(f"Failed to extract audio from video: {e}")