Informations détaillées

TranscriboZH is an open-source tool for transcribing audio and video files, developed by the Canton of Zurich. It uses the state-of-the-art Whisper v3 large model to deliver highly accurate transcriptions, including substantial support for Swiss German, without any licensing or usage fees. The application automatically identifies and distinguishes between different speakers in a recording (speaker diarization) and lets you define custom vocabulary for names or ambiguous terms. Transcripts can be edited directly in an integrated, browser-based editor that stays synchronized with the original audio or video, complete with keyboard shortcuts for efficient navigation. Results can be exported as plain text, SRT subtitle files, or as a standalone HTML file with synchronized playback that requires no installation to view. An optional summarization feature can generate concise summaries of transcripts using a locally run language model. Because the entire transcription pipeline runs on your own infrastructure, TranscriboZH is particularly well suited for processing confidential recordings without sending data to third-party cloud services.