itsskofficial/livewhisper-si-large-v3

si-large-v3 (CTranslate2)

A CTranslate2 conversion of kasunw/whisper-large-v3-sinhala for faster-whisper, as downloaded by LiveWhisper, a voice dictation app for Windows.

All credit for the model goes to its original authors. This repository only changes the file format (float16 weights, plus tokenizer.json and preprocessor_config.json that faster-whisper needs). The licence is the original's: apache-2.0.

The only model tried that writes Sinhala; still misses most words. Large-v3 sized, 3 GB converted.

Measured in LiveWhisper

metric	result
spoken Dakshina word error, native	77.6% (large-v3 114.7%)
spoken Dakshina character error, native	34.8% (large-v3 101.1%)
romanized, vs any accepted spelling	76.9% (large-v3 114.7%)

from faster_whisper import WhisperModel
model = WhisperModel("itsskofficial/livewhisper-si-large-v3", device="cuda", compute_type="int8_float16")