Open Mohand995 opened 1 month ago
You could also pass subtitles as an empty list to match_frames_and_subtitles
, inside the prepare_input
method, in minigpt4_video_inference.py
Make sure to comment out the call to extract_subtitles
since that fails for subtitle_path=None
The above changes would reduce the model to working only with the image frames.
In generate_prediction() function in minigpt4_video_inference.py file, instead of
if gen_subtitles
change it toif not gen_subtitles:
This will download openai's whisper model and generate subtitles.