Record 057 · Voice · 2023
67%23.57.V
faster-whisper
Guillaume Klein / SYSTRAN
“A CTranslate2 reimplementation that made local Whisper realtime on a CPU/GPU you own.”
faster-whisper is what a lot of HA, Frigate, and meeting tools actually call. Weights from OpenAI; speed from here.
Made local ASR practical, not batch-only.
- CTranslate2 Whisper
- Realtime-capable local ASR
- Quantized encoders
Whisper, fast enough to live-caption.
