Spaces:
Runtime error
Runtime error
Randriamiandry commited on
Commit ·
bdabebb
1
Parent(s): f7be0cf
Correction chargement Kokoro avec chemins explicites
Browse files
app.py
CHANGED
|
@@ -8,29 +8,47 @@ from fastapi.staticfiles import StaticFiles
|
|
| 8 |
from pydantic import BaseModel
|
| 9 |
import soundfile as sf
|
| 10 |
import time
|
|
|
|
| 11 |
|
| 12 |
print("🚀 Démarrage API Kokoro-82M (ONNX rapide)")
|
| 13 |
|
| 14 |
# -------------------------------------------------------------
|
| 15 |
-
#
|
| 16 |
# -------------------------------------------------------------
|
| 17 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 18 |
|
| 19 |
-
#
|
| 20 |
-
|
|
|
|
|
|
|
| 21 |
|
| 22 |
-
tts = Kokoro
|
| 23 |
print("✅ Modèle Kokoro chargé avec succès")
|
| 24 |
|
| 25 |
-
# Dictionnaire pour les voix clonées (
|
| 26 |
cloned_voices = {}
|
| 27 |
|
| 28 |
# -------------------------------------------------------------
|
| 29 |
# Application FastAPI
|
| 30 |
# -------------------------------------------------------------
|
| 31 |
-
app = FastAPI(title="Kokoro TTS API", version="2.
|
| 32 |
-
|
| 33 |
-
# Servir l'interface HTML (optionnel)
|
| 34 |
app.mount("/static", StaticFiles(directory="static"), name="static")
|
| 35 |
|
| 36 |
@app.get("/")
|
|
@@ -38,6 +56,7 @@ async def root():
|
|
| 38 |
return {
|
| 39 |
"status": "ok",
|
| 40 |
"model": "Kokoro-82M (ONNX)",
|
|
|
|
| 41 |
"endpoints": ["/tts", "/clone-voice", "/voices"]
|
| 42 |
}
|
| 43 |
|
|
@@ -47,7 +66,7 @@ async def demo():
|
|
| 47 |
|
| 48 |
class TTSRequest(BaseModel):
|
| 49 |
text: str
|
| 50 |
-
voice: str = "af_heart"
|
| 51 |
speed: float = 1.0
|
| 52 |
lang: str = "fr"
|
| 53 |
|
|
@@ -58,25 +77,22 @@ async def generate_speech(request: TTSRequest):
|
|
| 58 |
if not text:
|
| 59 |
raise HTTPException(400, "Texte vide")
|
| 60 |
|
| 61 |
-
# Choix de la voix : clonée ou intégrée
|
| 62 |
voice = request.voice
|
| 63 |
if voice in cloned_voices:
|
| 64 |
-
voice = cloned_voices[voice]
|
| 65 |
|
| 66 |
try:
|
| 67 |
-
# Génération audio
|
| 68 |
audio, sample_rate = tts.tts(
|
| 69 |
text=text,
|
| 70 |
voice=voice,
|
| 71 |
speed=request.speed,
|
| 72 |
lang=request.lang
|
| 73 |
)
|
| 74 |
-
# Sauvegarde en mémoire
|
| 75 |
buffer = io.BytesIO()
|
| 76 |
sf.write(buffer, audio, sample_rate, format='WAV')
|
| 77 |
buffer.seek(0)
|
| 78 |
|
| 79 |
-
print(f"✅ Généré en {time.time()-start:.2f}s | voix={voice}
|
| 80 |
return Response(content=buffer.read(), media_type="audio/wav")
|
| 81 |
except Exception as e:
|
| 82 |
raise HTTPException(500, f"Erreur TTS: {str(e)}")
|
|
@@ -86,21 +102,18 @@ async def clone_voice(audio_file: UploadFile = File(...), voice_name: str = Form
|
|
| 86 |
if voice_name in cloned_voices:
|
| 87 |
raise HTTPException(400, f"La voix '{voice_name}' existe déjà")
|
| 88 |
|
| 89 |
-
# Sauvegarde du fichier temporaire
|
| 90 |
with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp:
|
| 91 |
content = await audio_file.read()
|
| 92 |
tmp.write(content)
|
| 93 |
tmp_path = tmp.name
|
| 94 |
|
| 95 |
try:
|
| 96 |
-
# Certaines versions supportent clone_voice
|
| 97 |
if hasattr(tts, 'clone_voice'):
|
| 98 |
cloned_voice = tts.clone_voice(tmp_path)
|
|
|
|
|
|
|
| 99 |
else:
|
| 100 |
-
|
| 101 |
-
cloned_voice = tmp_path
|
| 102 |
-
cloned_voices[voice_name] = cloned_voice
|
| 103 |
-
return {"message": f"Voix '{voice_name}' clonée", "voice_name": voice_name, "total": len(cloned_voices)}
|
| 104 |
except Exception as e:
|
| 105 |
raise HTTPException(500, f"Erreur clonage: {str(e)}")
|
| 106 |
finally:
|
|
@@ -109,7 +122,6 @@ async def clone_voice(audio_file: UploadFile = File(...), voice_name: str = Form
|
|
| 109 |
|
| 110 |
@app.get("/voices")
|
| 111 |
async def list_voices():
|
| 112 |
-
# Les voix intégrées sont dans tts.voices (si dispo)
|
| 113 |
builtin = list(tts.voices.keys()) if hasattr(tts, 'voices') else []
|
| 114 |
return {"builtin": builtin, "cloned": list(cloned_voices.keys())}
|
| 115 |
|
|
|
|
| 8 |
from pydantic import BaseModel
|
| 9 |
import soundfile as sf
|
| 10 |
import time
|
| 11 |
+
from huggingface_hub import hf_hub_download
|
| 12 |
|
| 13 |
print("🚀 Démarrage API Kokoro-82M (ONNX rapide)")
|
| 14 |
|
| 15 |
# -------------------------------------------------------------
|
| 16 |
+
# Téléchargement des fichiers du modèle et des voix
|
| 17 |
# -------------------------------------------------------------
|
| 18 |
+
REPO_ID = "onnx-community/Kokoro-82M-v1.0-ONNX"
|
| 19 |
+
|
| 20 |
+
# Télécharger les deux fichiers nécessaires
|
| 21 |
+
try:
|
| 22 |
+
model_path = hf_hub_download(
|
| 23 |
+
repo_id=REPO_ID,
|
| 24 |
+
filename="model.onnx",
|
| 25 |
+
cache_dir="/app/cache"
|
| 26 |
+
)
|
| 27 |
+
voices_path = hf_hub_download(
|
| 28 |
+
repo_id=REPO_ID,
|
| 29 |
+
filename="voices.json",
|
| 30 |
+
cache_dir="/app/cache"
|
| 31 |
+
)
|
| 32 |
+
print(f"✅ Modèle téléchargé : {model_path}")
|
| 33 |
+
print(f"✅ Voices téléchargées : {voices_path}")
|
| 34 |
+
except Exception as e:
|
| 35 |
+
raise RuntimeError(f"Échec du téléchargement : {e}")
|
| 36 |
|
| 37 |
+
# -------------------------------------------------------------
|
| 38 |
+
# Chargement de Kokoro avec les chemins explicites
|
| 39 |
+
# -------------------------------------------------------------
|
| 40 |
+
from kokoro_onnx import Kokoro
|
| 41 |
|
| 42 |
+
tts = Kokoro(model_path=model_path, voices_path=voices_path)
|
| 43 |
print("✅ Modèle Kokoro chargé avec succès")
|
| 44 |
|
| 45 |
+
# Dictionnaire pour les voix clonées (optionnel)
|
| 46 |
cloned_voices = {}
|
| 47 |
|
| 48 |
# -------------------------------------------------------------
|
| 49 |
# Application FastAPI
|
| 50 |
# -------------------------------------------------------------
|
| 51 |
+
app = FastAPI(title="Kokoro TTS API", version="2.3")
|
|
|
|
|
|
|
| 52 |
app.mount("/static", StaticFiles(directory="static"), name="static")
|
| 53 |
|
| 54 |
@app.get("/")
|
|
|
|
| 56 |
return {
|
| 57 |
"status": "ok",
|
| 58 |
"model": "Kokoro-82M (ONNX)",
|
| 59 |
+
"voices_available": list(tts.voices.keys()) if hasattr(tts, 'voices') else [],
|
| 60 |
"endpoints": ["/tts", "/clone-voice", "/voices"]
|
| 61 |
}
|
| 62 |
|
|
|
|
| 66 |
|
| 67 |
class TTSRequest(BaseModel):
|
| 68 |
text: str
|
| 69 |
+
voice: str = "af_heart"
|
| 70 |
speed: float = 1.0
|
| 71 |
lang: str = "fr"
|
| 72 |
|
|
|
|
| 77 |
if not text:
|
| 78 |
raise HTTPException(400, "Texte vide")
|
| 79 |
|
|
|
|
| 80 |
voice = request.voice
|
| 81 |
if voice in cloned_voices:
|
| 82 |
+
voice = cloned_voices[voice]
|
| 83 |
|
| 84 |
try:
|
|
|
|
| 85 |
audio, sample_rate = tts.tts(
|
| 86 |
text=text,
|
| 87 |
voice=voice,
|
| 88 |
speed=request.speed,
|
| 89 |
lang=request.lang
|
| 90 |
)
|
|
|
|
| 91 |
buffer = io.BytesIO()
|
| 92 |
sf.write(buffer, audio, sample_rate, format='WAV')
|
| 93 |
buffer.seek(0)
|
| 94 |
|
| 95 |
+
print(f"✅ Généré en {time.time()-start:.2f}s | voix={voice}")
|
| 96 |
return Response(content=buffer.read(), media_type="audio/wav")
|
| 97 |
except Exception as e:
|
| 98 |
raise HTTPException(500, f"Erreur TTS: {str(e)}")
|
|
|
|
| 102 |
if voice_name in cloned_voices:
|
| 103 |
raise HTTPException(400, f"La voix '{voice_name}' existe déjà")
|
| 104 |
|
|
|
|
| 105 |
with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp:
|
| 106 |
content = await audio_file.read()
|
| 107 |
tmp.write(content)
|
| 108 |
tmp_path = tmp.name
|
| 109 |
|
| 110 |
try:
|
|
|
|
| 111 |
if hasattr(tts, 'clone_voice'):
|
| 112 |
cloned_voice = tts.clone_voice(tmp_path)
|
| 113 |
+
cloned_voices[voice_name] = cloned_voice
|
| 114 |
+
return {"message": f"Voix '{voice_name}' clonée", "voice_name": voice_name}
|
| 115 |
else:
|
| 116 |
+
raise HTTPException(501, "Clonage non supporté par cette version de kokoro-onnx")
|
|
|
|
|
|
|
|
|
|
| 117 |
except Exception as e:
|
| 118 |
raise HTTPException(500, f"Erreur clonage: {str(e)}")
|
| 119 |
finally:
|
|
|
|
| 122 |
|
| 123 |
@app.get("/voices")
|
| 124 |
async def list_voices():
|
|
|
|
| 125 |
builtin = list(tts.voices.keys()) if hasattr(tts, 'voices') else []
|
| 126 |
return {"builtin": builtin, "cloned": list(cloned_voices.keys())}
|
| 127 |
|