bn-medium (CTranslate2)

A CTranslate2 conversion of bengaliAI/tugstugi_bengaliai-asr_whisper-medium 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.

Winner of the Bengali.AI speech recognition competition. Half the size of large-v3 and faster, 1.5 GB once converted.

Measured in LiveWhisper

metric result
FLEURS word error 20.9% (large-v3 73.3%)
delivered Banglish 31.7% (large-v3 66.5%)
from faster_whisper import WhisperModel
model = WhisperModel("itsskofficial/livewhisper-bn-medium", device="cuda", compute_type="int8_float16")
Downloads last month
11
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for itsskofficial/livewhisper-bn-medium

Finetuned
(10)
this model