app.py

from transformers import pipeline
from transformers.utils import logging
import torch
import pandas as pd
import time
import gradio as gr

logging.set_verbosity_error()

asr = pipeline(task="automatic-speech-recognition",
              #  model="./models/distil-whisper/distil-small.en")
               model="distil-whisper/distil-small.en")

translator = pipeline(task="translation",
                      # model="./models/facebook/nllb-200-distilled-600M",
                      model="facebook/nllb-200-distilled-600M",
                      torch_dtype=torch.bfloat16)

flores_200_df = pd.read_csv("Flores200_language_codes.csv", encoding='cp1252')
flores_200 = dict(zip(flores_200_df['Language'],flores_200_df['FLORES_200_code']))
flores_200_languages = list(flores_200.keys())

def transcribe_audio(filepath, tgt_language):

    target_language = flores_200_df.loc[int(tgt_language),'Language']
    print(f"Selected Target Language: {target_language}")

    time.sleep(5)
    if filepath is None:
        gr.Warning("No audio found, please retry.")
        return ""

    english_transcript = asr(
      filepath,
      # max_new_tokens=256,
      chunk_length_s=30,
      batch_size=8,
    )['text']
    print(english_transcript)

    output = translator(english_transcript, src_lang="eng_Latn",
                        tgt_lang=flores_200_df.loc[int(tgt_language),'FLORES_200_code'])[0]['translation_text']

    print(output)
    return output

demo = gr.Blocks()

mic_transcribe = gr.Interface(
    fn=transcribe_audio,
    inputs=[gr.Audio(sources="microphone",
                    type="filepath"), gr.Dropdown(flores_200_df.Language.tolist(), type='index', label='Select Target Language')],
    outputs=gr.Textbox(label="Transcription in Selected Target Language",
                       lines=3),
    allow_flagging="never")

file_transcribe = gr.Interface(
    fn=transcribe_audio,
    inputs=[gr.Audio(sources="upload",
                    type="filepath"), gr.Dropdown(flores_200_df.Language.tolist(), type='index', label='Select Target Language')],
    outputs=gr.Textbox(label="Transcription in Selected Target Language",
                       lines=3),
    allow_flagging="never",
)

with demo:
    gr.TabbedInterface(
        [mic_transcribe,
         file_transcribe],
        ["Speak Through Microphone",
         "Upload Audio File"],
    )
demo.launch(debug=True)