app.py

import whisper
import gradio as gr
import time
import os

'''model = whisper.load_model("base")
print(model.device)'''


def speechtotext(tmp_filename, uploaded):
    try:
        source = uploaded if uploaded is not None else tmp_filename
        result = os.system("whisper" + source + " --language Hindi " + " --task translate ")
        return f'Detected language: {Language.make(language=result["language"]).display_name()}\n\n ' \
               f'You said: {result["text"]}'
    except:
        return "Unable to generate translation"


gr.Interface(

    title="NS-AI-Labs Custom Whisper",
    thumbnail="https://cdn.openai.com/whisper/asr-summary-of-model-architecture-desktop.svg",
    css="""
    .gr-prose p{text-align: center;}
    .gr-button {background: black;color: white}
    """,
    description="we customised whisper with some additional ASR layers , speak in any language we are here to get it "
                "recognised !",
    fn=speechtotext,
    inputs=[
        gr.Audio(label="Record your voice on your mic", source="microphone", type="filepath"),
        gr.Audio(source="upload", type="filepath", label="Upload Audio")],
    outputs="text").launch()