app.py

import gradio as gr
import torch
from transformers import pipeline

# Initialize the ASR pipeline for Bulgarian
asr_pipeline = pipeline("automatic-speech-recognition", model="infinitejoy/wav2vec2-large-xls-r-300m-bulgarian")


# ASR 변환 함수 (speech-to-text conversion)
def asr_generate(audio):
    transcription = asr_pipeline(audio)["text"]
    return transcription

# Gradio 인터페이스 생성
iface = gr.Interface(
    fn=asr_generate, 
    inputs=gr.Audio(source="microphone", type="filepath"), 
    outputs="text",
    title="Bulgarian Speech Recognition",
    description="Upload or record audio in Bulgarian to get the transcription."
)

# 인터페이스 실행
if __name__ == "__main__":
    iface.launch()