This is ivrit.ai's faster-whisper model, based on the ivrit-ai/whisper-v2-d4 Whisper model.

Training data includes 250 hours of volunteer-transcribed speech from the ivrit-ai/crowd-transcribe-v4 dataset, as well as 100 ours of professional transcribed speech from other sources.

Release date: September 8th, 2024.

Prerequisites

pip3 install faster_whisper

Usage

import faster_whisper
model = faster_whisper.WhisperModel('ivrit-ai/faster-whisper-v2-d4')

segs, _ = model.transcribe('media-file', language='he')

texts = [s.text for s in segs]

transcribed_text = ' '.join(texts)
print(f'Transcribed text: {transcribed_text}')