This tool accepts a local recording or a supported media URL. It prepares the audio, runs mlx-whisper and saves both a readable transcript and structured segments with timestamps.
Keeping the output useful
The Markdown transcript is convenient for reading and searching. The segment file retains timestamps for further processing. Both outputs can be moved into the project the recording belongs to.
Whisper provides the speech recognition. My work is the Python workflow around it, including media handling with yt-dlp and conversion with ffmpeg.
The tool supports Norwegian and English. Names and technical terms need checking, and it does not identify individual speakers.