whisper-large-v3-vaani-hindi-ggml
GGML conversion of ARTPARK-IISc/whisper-large-v3-vaani-hindi (a Whisper checkpoint) for whisper.cpp, in q4_0 / q5_0 / q8_0 quantizations (phones use q4/q5, Macs q8).
Source checkpoint: ARTPARK-IISc/whisper-large-v3-vaani-hindi by ARTPARK-IISc · License: apache-2.0 (unchanged; this repo only re-packages the weights)
Engine: Load with whisper.cpp (whisper-cli -m <file>), or any app that embeds it.
Files
| File | Quantization | Size | Note |
|---|---|---|---|
ggml-whisper-large-v3-vaani-hindi-q8_0.bin |
q8_0 | 1657 MB | |
ggml-whisper-large-v3-vaani-hindi-q5_0.bin |
q5_0 | 1081 MB | |
ggml-whisper-large-v3-vaani-hindi-q4_0.bin |
q4_0 | 889 MB |
f16 is the lossless conversion; q8_0 is nearly identical in accuracy at ~55 % of the size; q5_0/q5_k are the phone-friendly choice; q4_* is smallest with a small accuracy cost.
How these were made
Converted from the upstream checkpoint with the engine's own converter, then quantized with the engine's quantizer. Each variant was checked by transcribing short Portuguese and English samples before upload.
Attribution
Weights are derivative works of the upstream model and keep its license. Please cite the original authors (ARTPARK-IISc). The conversion and hosting here are maintained by JoaoZaokk so that the download links used by the Odysseus / Open WebUI native apps stay stable. No warranty.