Transcription Stream

Transcription Stream turns recordings into transcripts with speaker labels for people who need to review audio on their own hardware. It works fully offline, including summaries through Ollama and Mistral.

• Transcription and speaker labels: Submit recordings through the browser or SSH to produce text and identify who spoke when.
• Audio review: Follow synchronized transcript highlighting and scrolling, jump within the recording through the text, and adjust playback speed.
• AI summaries: Generate summaries with Ollama and Mistral, with a customizable prompt to control their focus.
• Transcript search: Meilisearch provides full-text search across transcriptions.
• Result access: Download output files from the web interface and follow task completion alerts to the results.

The Docker deployment requires an NVIDIA GPU. Ollama can connect to a different endpoint, so summary processing can run on another machine rather than alongside transcription.

The project is open source under GPL-3.0. The supplied deployment is example code and requires additional security measures before production use.