Repository navigation
Releases
Releases · incordlabs/QORA-STT-TINY
QORA-STT v0.1.0 - Pure Rust Speech-to-Text
QORA-STT v0.1.0
Pure Rust speech-to-text engine based on OpenAI Whisper Tiny (39M params).
Features
- 99 languages supported
- F32 weights (no quantization needed at 39M params)
- ~3.5s for 6 seconds of audio on i5-11500
- CPU-only, no GPU required
Download & Run
Download all 4 files into the same directory:
qora-stt.exe --model-path . --load model.qora-stt --audio recording.wavFiles
| File | Size | Description |
|---|---|---|
model.qora-stt |
144 MB | F32 model weights |
qora-stt.exe |
2.5 MB | Windows inference engine |
tokenizer.json |
2.4 MB | BPE tokenizer (51,865 vocab) |
config.json |
2.0 KB | Model configuration |
Full Changelog: https://github.com/qora-protocol/QORA-STT-TINY/commits/v0.1.0