main.py

from fastapi import FastAPI, UploadFile, File
from transformers import pipeline
import librosa
from deep_translator import GoogleTranslator
import io

app = FastAPI()

print("Loading Speech Recognition")
pipe = pipeline("automatic-speech-recognition", model="Akashpb13/xlsr_kurmanji_kurdish")
print("Speech Recognition Loaded")

print("Loading translator")
translator = GoogleTranslator(source='ku', target='fr')
print("Translator loaded")


def speech2text(audio_data: bytes):
    audio_array, _ = librosa.load(io.BytesIO(audio_data), sr=16000)    
    output = pipe(audio_array)
    return output["text"]

@app.post("/transcribe")
async def transcribe(file: UploadFile = File(...)):
    audio_data = await file.read()
    text_output = speech2text(audio_data)
    translated = translator.translate(text_output)
    return {"text": text_output, "translation": translated}