itsskofficial/livewhisper-sd-large

sd-large (CTranslate2)

A CTranslate2 conversion of steja/whisper-large-sindhi 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.

large-v3 cannot write Sindhi at all; this can. Large-v2 sized, allow ~6 GB of VRAM with the main model.

Measured in LiveWhisper

metric	result
FLEURS test word error, native	29.2% (large-v3 104%)
FLEURS test character error, native	13.1% (large-v3 108%)

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