Turn interviews, focus groups, and field recordings into accurate text transcripts in minutes.
Whisper
Calame: Transcription pipeline for all
Tools for research and application workflow from raw audio to clean, structured transcripts.
Why Calame
The transcription pipeline
at the state of the art
Calame turns your interviews and field recordings into clean, structured transcripts with diarization, anonymisation, and full privacy.
Automatically identify and label each speaker so you always know who said what.
Pyannote
Calame detects and redacts names, places, and identifying details to keep your participants protected.
Stanza
All tools are open-source and available on 
Hugging Face
Performance
Fast transcriptionon your hardware
| Hardware | File | tTRS | tDIA | t |
|---|---|---|---|---|
| i7-1260P | 5 min | 5.51 | 3.41 | 8.92 |
| 30 min | 26.34 | 22.72 | 49.06 | |
| 60 min | 51.31 | 39.45 | 90.76 | |
| RTX 2070 | 5 min | 3.24 | 0.39 | 3.63 |
| 30 min | 24.99 | 4.30 | 29.29 | |
| 60 min | 38.32 | 13.24 | 51.56 | |
| RTX 4060 | 5 min | 1.02 | 0.24 | 1.26 |
| 30 min | 3.72 | 1.36 | 5.08 | |
| 60 min | 6.99 | 2.83 | 9.82 |
System Requirements
Built to runon almost any machine
12 GB
RAM
GPU / CPU
Supported
Docker
Environment
Future directions
Current research anddevelopment focus
Under-resourced languages
Expanding first-class support beyond Québécois French to more dialects and low-resource languages.
Targeted diarization
Focus speaker identification on a single participant versus the full group, for interviews and one-on-one recordings.
Multi-model usage
Choose from multiple transcription or speaker separation models to improve accuracy across diverse audio conditions.
User collaboration
Share projects, review transcripts together, and manage team access for collaborative research workflows.
Our Partners
