Whishper
An open-source speech-to-text web app that transcribes and translates locally with Whisper models. Works offline on CPU or NVIDIA GPU under AGPL-3.0.
Whishper is a self-hosted speech-to-text app for people who need transcripts or translated subtitles from audio and video. Its browser interface brings transcription, translation and subtitle editing together, with all three running on your own machine. It can work offline, so local media doesn't need to go to a cloud transcription service.
You can upload media files or import them from URLs supported by yt-dlp. URL imports need access to the source; offline use applies to media you already have locally. Whishper uses Whisper models through FasterWhisper and can run on a CPU, with NVIDIA GPU support for faster transcription.
The subtitle editor pairs text with media playback and highlights the current segment as the recording plays. You can split segments, insert new ones and select the subtitle language. Characters-per-second warnings help identify captions that may be too dense to read in the time available.
Translation uses LibreTranslate locally and covers the languages it supports. For people preparing subtitles in another language, that keeps translation and subsequent edits in the same interface. Finished transcripts can leave the app as plain TXT or structured JSON, while SRT and VTT exports suit subtitle workflows. You can also copy the raw transcript text.
Whishper is open source under the GNU Affero General Public License v3.0 (AGPL-3.0). Its local components include a transcription service, LibreTranslate for translation and MongoDB for storing transcription information.
