app.py

from flask import Flask, request, jsonify
from transformers import AutoProcessor, SeamlessM4Tv2Model
import numpy as np
import sounddevice as sd
import wave
import os
from huggingface_hub import InferenceClient, login
from dotenv import load_dotenv
app = Flask(__name__)

load_dotenv()
hftoken = os.getenv("HF_TOKEN")



login(token=hftoken)

processor = AutoProcessor.from_pretrained("facebook/seamless-m4t-v2-large" )
model = SeamlessM4Tv2Model.from_pretrained("facebook/seamless-m4t-v2-large")

UPLOAD_FOLDER = "audio_files"
os.makedirs(UPLOAD_FOLDER, exist_ok=True)

@app.route("/", methods=["GET"])
def return_text():
    return jsonify({"text": "Hello, world!"})

@app.route("/record", methods=["POST"])
def record_audio():
    file = request.files['audio']
    filename = os.path.join(UPLOAD_FOLDER, file.filename)
    file.save(filename)
    
    # Charger et traiter l'audio
    audio_data, orig_freq = torchaudio.load(filename)
    audio_inputs = processor(audios=audio_data, return_tensors="pt")
    output_tokens = model.generate(**audio_inputs, tgt_lang="fra", generate_speech=False)
    translated_text = processor.decode(output_tokens[0].tolist()[0], skip_special_tokens=True)
    
    return jsonify({"translated_text": translated_text})

@app.route("/text_to_speech", methods=["POST"])
def text_to_speech():
    data = request.get_json()
    text = data.get("text")
    src_lang = data.get("src_lang")
    tgt_lang = data.get("tgt_lang")
    
    text_inputs = processor(text=text, src_lang=src_lang, return_tensors="pt")
    audio_array = model.generate(**text_inputs, tgt_lang=tgt_lang)[0].cpu().numpy().squeeze()
    
    output_filename = os.path.join(UPLOAD_FOLDER, "output.wav")
    with wave.open(output_filename, "wb") as wf:
        wf.setnchannels(1)
        wf.setsampwidth(2)
        wf.setframerate(16000)
        wf.writeframes((audio_array * 32767).astype(np.int16).tobytes())
    
    return jsonify({"audio_url": output_filename})

if __name__ == "__main__":
    app.run(debug=True)