Skip to content

[KDD'2026] "VideoRAG: Chat with Your Videos"

update support to non-audio videos

Re-bin6bba4649f05f
diff --git a/VideoRAG-algorithm/videorag/_videoutil/asr.py b/VideoRAG-algorithm/videorag/_videoutil/asr.py
index 820bf1d..b29b128 100755
--- a/VideoRAG-algorithm/videorag/_videoutil/asr.py
+++ b/VideoRAG-algorithm/videorag/_videoutil/asr.py
@@ -15,6 +15,12 @@ def speech_to_text(video_name, working_dir, segment_index2name, audio_output_for
for index in tqdm(segment_index2name, desc=f"Speech Recognition {video_name}"):
segment_name = segment_index2name[index]
audio_file = os.path.join(cache_path, f"{segment_name}.{audio_output_format}")
+
+ # if the audio file does not exist, skip it
+ if not os.path.exists(audio_file):
+ transcripts[index] = ""
+ continue
+
segments, info = model.transcribe(audio_file)
result = ""
for segment in segments:
diff --git a/VideoRAG-algorithm/videorag/_videoutil/split.py b/VideoRAG-algorithm/videorag/_videoutil/split.py
index e0c1252..41d4367 100755
--- a/VideoRAG-algorithm/videorag/_videoutil/split.py
+++ b/VideoRAG-algorithm/videorag/_videoutil/split.py
@@ -5,6 +5,7 @@ import numpy as np
from tqdm import tqdm
from moviepy.video import fx as vfx
from moviepy.video.io.VideoFileClip import VideoFileClip
+from .._utils import logger
def split_video(
video_path,
@@ -47,9 +48,12 @@ def split_video(
# save audio
audio_file_base_name = segment_index2name[f"{segment_index}"]
audio_file = f'{audio_file_base_name}.{audio_output_format}'
- subaudio = subvideo.audio
- subaudio.write_audiofile(os.path.join(video_segment_cache_path, audio_file), codec='mp3', verbose=False, logger=None)
-
+ try:
+ subaudio = subvideo.audio
+ subaudio.write_audiofile(os.path.join(video_segment_cache_path, audio_file), codec='mp3', verbose=False, logger=None)
+ except Exception as e:
+ logger.warning(f"Warning: Failed to extract audio for video {video_name} ({start}-{end}). Probably due to lack of audio track.")
+
segment_index += 1
return segment_index2name, segment_times_info