Whisper Large V3 Turbo Russian

🚀 Whisper Large V3 Turbo Russian Model

This is an automatic speech recognition model trained on Russian data, aiming to provide high - accuracy speech - to - text conversion.

🚀 Quick Start

This model is an automatic speech recognition model. It was trained with specific hardware configurations and a particular dataset. You can use the provided Python code example to quickly start using this model for speech transcription.

✨ Features

• Trained on the mozilla - foundation/common_voice_17_0 dataset, which contains rich Russian speech data.
• Based on the openai/whisper - large - v3 - turbo base model, leveraging its powerful pre - trained capabilities.
• Achieves high - accuracy speech recognition as measured by the accuracy metric.

📦 Installation

The README does not provide specific installation steps, so this section is skipped.

💻 Usage Examples

Basic Usage

from transformers import pipeline
import gradio as gr
import time

pipe = pipeline(
    model="dvislobokov/whisper-large-v3-turbo-russian",
    tokenizer="dvislobokov/whisper-large-v3-turbo-russian",
    task='automatic-speech-recognition',
    device='cpu'
)

def transcribe(audio):
    start = time.time()
    text = pipe(audio, return_timestamps=True)['text']
    print(time.time() - start)
    return text

iface = gr.Interface(
    fn=transcribe,
    inputs=gr.Audio(sources=['microphone', 'upload'], type='filepath'),
    outputs='text'
)

iface.launch(share=True)

Advanced Usage

The README does not provide advanced usage examples, so this part is skipped.

🔧 Technical Details

This model was trained using two A100 GPUs with 40 GB each, 128 GB of RAM, and 2 x Xeon 48 - Core 2.4 GHz CPUs. Approximately 7 hours were spent on training, and the training dataset consisted of 118k audio samples from Mozilla Common Voice 17.

📄 License

This project is licensed under the MIT License.

📚 Documentation