Dravidian Whisper language packs (ggml q5_0)
On-demand language packs for the Smritsavant voice engine: ggml q5_0 conversions of community Whisper fine-tunes. Conversion pipeline: HF pytorch_model.bin → ggml f16 (whisper.cpp convert-h5-to-ggml.py) → q5_0 (whisper-quantize). Base architecture: openai/whisper-medium. All upstream models are Apache-2.0.
| File | Language | Upstream model | Author |
|---|---|---|---|
| ggml-tamil-medium-q5.bin | Tamil | vasista22/whisper-tamil-medium | Speech Lab, IIT Madras (Bhashini) |
| ggml-te-medium-q5.bin | Telugu | vasista22/whisper-telugu-medium | Speech Lab, IIT Madras (Bhashini) |
| ggml-kn-medium-q5.bin | Kannada | vasista22/whisper-kannada-medium | Speech Lab, IIT Madras (Bhashini) |
| ggml-ml-medium-q5.bin | Malayalam | thennal/whisper-medium-ml | thennal |