faster-whisper
A reimplementation of Whisper that transcribes several times faster on the same hardware. Local, MIT, and the sensible default.
- Self-hosted
- Allowed
- No GPU needed
Whisper large-v3
Speech to text, MIT-licensed, running entirely on your machine. The one uncomplicated thing on this page.
- Open weights
- Audio
- Allowed adult content
At a glance
- Made by
- OpenAI
- Family
- Whisper
- Licence
- MIT
- Size
- 1.55B parameters
- Graphics memory
- 6 GB minimum, 10 GB recommended
- Adult content
- AllowedAdult content is permitted within the published rules.
- On Hugging Face
- openai/whisper-large-v3
An MIT licence, weights you can download, and no telemetry: transcription is the one part of this field where the free option is also the private one and also the best one.
The large model wants about 10 GB of VRAM; the smaller sizes run on almost anything, and faster-whisper makes even the large one practical on a modest card.