Create app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,76 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
from fastapi import FastAPI, Response, HTTPException
|
| 2 |
+
from fastapi.middleware.cors import CORSMiddleware
|
| 3 |
+
from melo.api import TTS
|
| 4 |
+
import os
|
| 5 |
+
|
| 6 |
+
app = FastAPI()
|
| 7 |
+
|
| 8 |
+
app.add_middleware(
|
| 9 |
+
CORSMiddleware,
|
| 10 |
+
allow_origins=["*"],
|
| 11 |
+
allow_methods=["*"],
|
| 12 |
+
allow_headers=["*"],
|
| 13 |
+
)
|
| 14 |
+
|
| 15 |
+
# Stockage des modèles chargés
|
| 16 |
+
models = {}
|
| 17 |
+
device = 'cpu'
|
| 18 |
+
|
| 19 |
+
print("⏳ Initialisation de MeloTTS...")
|
| 20 |
+
|
| 21 |
+
try:
|
| 22 |
+
# On pré-charge le Français et l'Anglais
|
| 23 |
+
# C'est rapide et léger avec Melo
|
| 24 |
+
print(" - Chargement FR...")
|
| 25 |
+
models['FR'] = TTS(language='FR', device=device)
|
| 26 |
+
|
| 27 |
+
print(" - Chargement EN...")
|
| 28 |
+
models['EN'] = TTS(language='EN', device=device)
|
| 29 |
+
|
| 30 |
+
print("✅ MeloTTS Prêt (FR + EN) !")
|
| 31 |
+
except Exception as e:
|
| 32 |
+
print(f"❌ Erreur au chargement : {e}")
|
| 33 |
+
|
| 34 |
+
@app.post("/tts")
|
| 35 |
+
async def generate_speech(data: dict):
|
| 36 |
+
text = data.get("text", "")
|
| 37 |
+
lang = data.get("lang", "FR").upper() # Par défaut FR
|
| 38 |
+
speed = data.get("speed", 1.0)
|
| 39 |
+
|
| 40 |
+
if not text:
|
| 41 |
+
raise HTTPException(status_code=400, detail="Texte vide")
|
| 42 |
+
|
| 43 |
+
# Vérification si la langue est chargée
|
| 44 |
+
if lang not in models:
|
| 45 |
+
# Tentative de chargement dynamique (si vous demandez ES ou ZH par exemple)
|
| 46 |
+
try:
|
| 47 |
+
print(f"⏳ Chargement dynamique de {lang}...")
|
| 48 |
+
models[lang] = TTS(language=lang, device=device)
|
| 49 |
+
except Exception as e:
|
| 50 |
+
raise HTTPException(status_code=400, detail=f"Langue {lang} non supportée ou erreur: {e}")
|
| 51 |
+
|
| 52 |
+
model = models[lang]
|
| 53 |
+
speaker_ids = model.hps.data.spk2id
|
| 54 |
+
|
| 55 |
+
# MeloTTS utilise souvent le code langue comme ID speaker (ex: 'FR')
|
| 56 |
+
# Mais parfois c'est 'EN-US', 'EN-BR', etc. pour l'anglais.
|
| 57 |
+
# On prend le premier dispo si la clé exacte n'existe pas.
|
| 58 |
+
spk_id = speaker_ids.get(lang, list(speaker_ids.values())[0])
|
| 59 |
+
|
| 60 |
+
output_path = f"/tmp/output_{os.getpid()}.wav"
|
| 61 |
+
|
| 62 |
+
try:
|
| 63 |
+
model.tts_to_file(text, spk_id, output_path, speed=speed)
|
| 64 |
+
|
| 65 |
+
with open(output_path, "rb") as f:
|
| 66 |
+
audio_data = f.read()
|
| 67 |
+
|
| 68 |
+
return Response(content=audio_data, media_type="audio/wav")
|
| 69 |
+
|
| 70 |
+
except Exception as e:
|
| 71 |
+
print(f"❌ Erreur génération : {e}")
|
| 72 |
+
return Response(content=str(e), status_code=500)
|
| 73 |
+
|
| 74 |
+
@app.get("/")
|
| 75 |
+
def home():
|
| 76 |
+
return {"status": "MeloTTS API Ready", "loaded_langs": list(models.keys())}
|