Whishper
A self-hosted web application that transcribes audio or video to text, translates it, and edits subtitles entirely on your own machine using Whisper models.
Whishper transcribes audio and video, including from any source yt-dlp supports or an uploaded file, using FasterWhisper as its backend, and can run on either GPU or CPU. Transcriptions can be downloaded as TXT, JSON, VTT, or SRT, or copied as raw text, and can be translated into any language supported by LibreTranslate.
A built-in subtitle editor highlights the transcription in sync with media playback, warns about characters-per-second limits, and supports splitting and inserting segments without leaving the interface. The project is composed of a transcription API built on FasterWhisper, a backend that coordinates the frontend, database, and background tasks, a SvelteKit frontend, and third-party MongoDB, LibreTranslate, and Nginx containers, all running entirely locally. It is licensed under AGPL-3.0, and the maintainer notes a complete rewrite is in progress on a separate branch.