yt-whisper is a small, focused command-line tool that automatically generates subtitles for YouTube videos using OpenAI’s Whisper speech recognition model. The open-source project has around 1,400 stars on GitHub and does one job well: give it a video URL, get back a subtitle file.
The tool chains together proven components: it downloads the audio track of a YouTube video with yt-dlp, converts it as needed with ffmpeg, and runs it through Whisper to produce a timestamped transcript saved as a standard subtitle file you can upload to YouTube or use anywhere else. You can choose among Whisper model sizes to trade speed against accuracy, and transcription runs locally on your machine, so there are no API fees and no audio leaves your computer. It is especially handy for creators subtitling their back catalog, viewers who need captions on videos that lack them, and anyone building transcript datasets from public videos. The codebase is compact and readable, which has also made it a popular starting point for people writing their own transcription scripts. Its most recent updates were in early 2024, reflecting a finished, stable utility rather than an abandoned one. It is MIT-licensed.
Free. yt-whisper is fully open source under the MIT license and runs on your own hardware at no cost.
Creators, researchers, and viewers who need quick, accurate subtitles or transcripts for YouTube videos without paying for a transcription service.