bugbakery/transcribee

What it solves

Transcribee simplifies the process of converting audio and video files into text. It addresses the need for a faster, more accessible workflow that combines automatic transcription with easy manual correction and real-time collaboration.

How it works

The software follows a three-step process:

1. Import: Media files are automatically converted to text using whisper.cpp for transcription, Wav2Vec2 for realignment, and speechbrain for speaker identification.
2. Improvement: Users can manually edit the draft transcript to fix errors and assign names to speakers. This process is collaborative, using the automerge CRDT to sync changes instantly between multiple users.
3. Export: The final corrected transcript is exported for use.

Who it’s for

It is designed for anyone needing to transcribe media files, whether individually or as part of a team collaborating on a single document.

Highlights

• Automatic Drafts: Generates initial transcripts using state-of-the-art AI models.
• Collaborative Editing: Allows multiple people to work on the same transcript simultaneously.
• Speaker Identification: Automatically detects different speakers in the audio file.
• Open Source: The project is open-source and can be self-hosted for data privacy.