app.py

import gradio as gr

import whisper
from transformers import MarianMTModel, MarianTokenizer, pipeline

model_name = "Helsinki-NLP/opus-mt-tc-big-lt-en"
tokenizer = MarianTokenizer.from_pretrained(model_name)
translation_model = MarianMTModel.from_pretrained(model_name)

model_name = "Aismantas/whisper-base-lithuanian"
asr_pipeline = pipeline("automatic-speech-recognition", model=model_name)

def transcribe(filepath):
    # Assuming the file is named 'audio.wav'
    # Run the transcription
    transcript = asr_pipeline(filepath)
    return translation_model.generate(**tokenizer(transcript['text'], return_tensors="pt", padding=True))


demo = gr.Interface(fn=transcribe, inputs=[gr.Audio(type='filepath')], outputs="text")
demo.launch()