ortexsolution/VR1

1import torch2 3import gradio as gr4from transformers import pipeline5 6model_ids = {7"Fast": "openai/whisper-small",8"Balanced": "openai/whisper-medium",9"Accurate": "openai/whisper-large-v3-turbo"10}11 12device = 0 if torch.cuda.is_available() else "cpu"13 14def transcribe(inputs, model_name):15    if inputs is None:16        raise gr.Error("No audio file submitted!")17 18    pipe = pipeline(19        task="automatic-speech-recognition",20        model=model_ids[model_name],21        chunk_length_s=30,22        device=device,23    )24    text = pipe(inputs, batch_size=8, generate_kwargs={"task": "transcribe"}, return_timestamps=True)["text"] 25    return  text26 27demo = gr.Blocks(theme=gr.themes.Ocean())28 29file_transcribe = gr.Interface(30    fn=transcribe,31    inputs=[32        gr.Audio(sources="upload", type="filepath", label="Audio file"),33        gr.Radio(list(model_ids.keys()), label="Select Model", value="Fast")34    ],35    outputs="text",36    title="Automated transcription of free-form voice comments",37    description=(38        "Please upload your audio file first, then select your model. Finally, click the submit button."39    ),40    flagging_mode="never",41)42mf_transcribe = gr.Interface(43    fn=transcribe,44    inputs=[45        gr.Audio(sources="microphone", type="filepath"),46        gr.Radio(list(model_ids.keys()), label="Select Model", value="Fast")47    ],48    outputs="text",49    title="Automated transcription of free-form voice comments",50    description=(51        "Please Record your audio first, then select your model. Finally, click the submit button."52    ),53    flagging_mode="never",54)55 56with demo:57    gr.TabbedInterface([file_transcribe,  mf_transcribe], ["Audio File","Microphone"])58 59demo.queue().launch(ssr_mode=False)60 61