Update app.py
Browse files
app.py
CHANGED
|
@@ -14,15 +14,14 @@ AUDIO_DIR = "/tmp/audio_output"
|
|
| 14 |
os.makedirs(TEMP_DIR, exist_ok=True)
|
| 15 |
os.makedirs(AUDIO_DIR, exist_ok=True)
|
| 16 |
|
| 17 |
-
# Paksa semua unduhan model masuk ke /tmp
|
| 18 |
os.environ["TTS_HOME"] = TEMP_DIR
|
| 19 |
-
|
| 20 |
-
# Persetujuan Lisensi Otomatis
|
| 21 |
os.environ["COQUI_TOS_AGREED"] = "1"
|
| 22 |
|
| 23 |
app = FastAPI(
|
| 24 |
-
title="PasBlast XTTS-v2 API",
|
| 25 |
-
description="API
|
|
|
|
| 26 |
)
|
| 27 |
security = HTTPBasic()
|
| 28 |
|
|
@@ -34,7 +33,7 @@ def get_xtts_instance() -> TTS:
|
|
| 34 |
try:
|
| 35 |
print("Memuat model XTTS-v2 ke RAM...")
|
| 36 |
tts_model = TTS(model_name="tts_models/multilingual/multi-dataset/xtts_v2", gpu=False)
|
| 37 |
-
print("Model berhasil dimuat
|
| 38 |
except Exception as e:
|
| 39 |
error_trace = traceback.format_exc()
|
| 40 |
print(f"Error Load Model: {error_trace}")
|
|
@@ -55,58 +54,96 @@ def verify_auth(credentials: HTTPBasicCredentials = Depends(security)):
|
|
| 55 |
|
| 56 |
# --- ENDPOINTS ---
|
| 57 |
|
| 58 |
-
@app.get("/")
|
| 59 |
def root():
|
| 60 |
-
return {
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 61 |
|
| 62 |
-
@app.get("/speakers", tags=["Info"])
|
| 63 |
def list_speakers(username: str = Depends(verify_auth)):
|
| 64 |
-
"""Melihat daftar karakter suara bawaan XTTS-v2."""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 65 |
tts = get_xtts_instance()
|
| 66 |
-
|
|
|
|
|
|
|
| 67 |
|
| 68 |
-
@app.post("/tts", tags=["Generation"])
|
| 69 |
def generate_tts(
|
| 70 |
-
text: str = Form(..., description="Teks
|
| 71 |
-
speaker: str = Form("Ana Florence", description="
|
| 72 |
-
language: str = Form("id", description="
|
|
|
|
|
|
|
|
|
|
| 73 |
username: str = Depends(verify_auth)
|
| 74 |
):
|
| 75 |
-
"""Sintesis teks
|
| 76 |
tts = get_xtts_instance()
|
| 77 |
output_path = os.path.join(AUDIO_DIR, f"tts_{uuid.uuid4().hex}.wav")
|
| 78 |
|
| 79 |
-
# 1. Validasi Speaker (Pencegah Error 500)
|
| 80 |
available_speakers = tts.speakers
|
| 81 |
if speaker not in available_speakers:
|
| 82 |
-
print(f"
|
| 83 |
speaker = available_speakers[0]
|
| 84 |
|
| 85 |
try:
|
| 86 |
-
print(f"
|
| 87 |
-
# 2. split_sentences=True wajib aktif untuk mencegah crash pada teks panjang
|
| 88 |
tts.tts_to_file(
|
| 89 |
text=text,
|
| 90 |
speaker=speaker,
|
| 91 |
language=language,
|
| 92 |
file_path=output_path,
|
| 93 |
-
split_sentences=True
|
|
|
|
|
|
|
|
|
|
| 94 |
)
|
| 95 |
-
|
| 96 |
-
return FileResponse(output_path, media_type="audio/wav", filename="pasblast_normal.wav")
|
| 97 |
except Exception as e:
|
| 98 |
error_trace = traceback.format_exc()
|
| 99 |
-
print(f"FATAL ERROR
|
| 100 |
-
raise HTTPException(status_code=500, detail=f"Gagal memproses TTS: {str(e)}
|
| 101 |
|
| 102 |
-
@app.post("/tts_voice_clone", tags=["Generation"])
|
| 103 |
def generate_tts_voice_clone(
|
| 104 |
-
text: str = Form(...),
|
| 105 |
-
language: str = Form("id"),
|
| 106 |
-
reference_audio: UploadFile = File(..., description="File WAV suara target (5-10 detik)"),
|
|
|
|
| 107 |
username: str = Depends(verify_auth)
|
| 108 |
):
|
| 109 |
-
"""
|
| 110 |
tts = get_xtts_instance()
|
| 111 |
|
| 112 |
ref_path = os.path.join(AUDIO_DIR, f"ref_{uuid.uuid4().hex}.wav")
|
|
@@ -116,16 +153,51 @@ def generate_tts_voice_clone(
|
|
| 116 |
output_path = os.path.join(AUDIO_DIR, f"clone_{uuid.uuid4().hex}.wav")
|
| 117 |
|
| 118 |
try:
|
| 119 |
-
print("
|
| 120 |
tts.tts_to_file(
|
| 121 |
text=text,
|
| 122 |
language=language,
|
| 123 |
speaker_wav=ref_path,
|
| 124 |
file_path=output_path,
|
| 125 |
-
split_sentences=True
|
|
|
|
| 126 |
)
|
| 127 |
return FileResponse(output_path, media_type="audio/wav", filename="pasblast_cloned.wav")
|
| 128 |
except Exception as e:
|
| 129 |
error_trace = traceback.format_exc()
|
| 130 |
-
print(f"FATAL ERROR
|
| 131 |
-
raise HTTPException(status_code=500, detail=f"Gagal
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 14 |
os.makedirs(TEMP_DIR, exist_ok=True)
|
| 15 |
os.makedirs(AUDIO_DIR, exist_ok=True)
|
| 16 |
|
| 17 |
+
# Paksa semua unduhan model masuk ke /tmp agar sesuai dengan penyimpanan ephemeral
|
| 18 |
os.environ["TTS_HOME"] = TEMP_DIR
|
|
|
|
|
|
|
| 19 |
os.environ["COQUI_TOS_AGREED"] = "1"
|
| 20 |
|
| 21 |
app = FastAPI(
|
| 22 |
+
title="PasBlast Comprehensive Coqui XTTS-v2 API",
|
| 23 |
+
description="API Komplet untuk seluruh kapabilitas Coqui XTTS-v2 yang dimungkinkan berjalan di CPU Space.",
|
| 24 |
+
version="1.0.0"
|
| 25 |
)
|
| 26 |
security = HTTPBasic()
|
| 27 |
|
|
|
|
| 33 |
try:
|
| 34 |
print("Memuat model XTTS-v2 ke RAM...")
|
| 35 |
tts_model = TTS(model_name="tts_models/multilingual/multi-dataset/xtts_v2", gpu=False)
|
| 36 |
+
print("Model berhasil dimuat!")
|
| 37 |
except Exception as e:
|
| 38 |
error_trace = traceback.format_exc()
|
| 39 |
print(f"Error Load Model: {error_trace}")
|
|
|
|
| 54 |
|
| 55 |
# --- ENDPOINTS ---
|
| 56 |
|
| 57 |
+
@app.get("/", tags=["Status"])
|
| 58 |
def root():
|
| 59 |
+
return {
|
| 60 |
+
"status": "online",
|
| 61 |
+
"message": "XTTS-v2 API Komplet Aktif. Akses /docs untuk Swagger UI Interaktif.",
|
| 62 |
+
"storage_info": {
|
| 63 |
+
"temp_dir": TEMP_DIR,
|
| 64 |
+
"audio_dir": AUDIO_DIR
|
| 65 |
+
}
|
| 66 |
+
}
|
| 67 |
+
|
| 68 |
+
@app.get("/models", tags=["Metadata & Info"])
|
| 69 |
+
def list_models(username: str = Depends(verify_auth)):
|
| 70 |
+
"""Melihat daftar seluruh model yang terdaftar secara internal di library Coqui TTS."""
|
| 71 |
+
try:
|
| 72 |
+
all_models = TTS.list_models()
|
| 73 |
+
return {"total_available_models": len(all_models), "models": all_models}
|
| 74 |
+
except Exception as e:
|
| 75 |
+
raise HTTPException(status_code=500, detail=f"Gagal mengambil list model: {str(e)}")
|
| 76 |
+
|
| 77 |
+
@app.get("/current_model", tags=["Metadata & Info"])
|
| 78 |
+
def get_current_model_status(username: str = Depends(verify_auth)):
|
| 79 |
+
"""Mengecek status dan informasi model yang saat ini sedang aktif di memori."""
|
| 80 |
+
global tts_model
|
| 81 |
+
return {
|
| 82 |
+
"loaded_in_memory": tts_model is not None,
|
| 83 |
+
"active_model_name": "tts_models/multilingual/multi-dataset/xtts_v2" if tts_model else None,
|
| 84 |
+
"device": "cpu"
|
| 85 |
+
}
|
| 86 |
|
| 87 |
+
@app.get("/speakers", tags=["Metadata & Info"])
|
| 88 |
def list_speakers(username: str = Depends(verify_auth)):
|
| 89 |
+
"""Melihat daftar seluruh karakter suara (speaker) bawaan yang didukung oleh XTTS-v2."""
|
| 90 |
+
tts = get_xtts_instance()
|
| 91 |
+
return {"total_speakers": len(tts.speakers), "speakers": tts.speakers}
|
| 92 |
+
|
| 93 |
+
@app.get("/languages", tags=["Metadata & Info"])
|
| 94 |
+
def list_languages(username: str = Depends(verify_auth)):
|
| 95 |
+
"""Melihat daftar kode bahasa (multilingual) yang didukung resmi oleh XTTS-v2."""
|
| 96 |
tts = get_xtts_instance()
|
| 97 |
+
# XTTS-v2 mendukung bahasa berkode resmi seperti 'id', 'en', 'es', dll.
|
| 98 |
+
languages = tts.languages if hasattr(tts, "languages") else ["id", "en", "es", "fr", "de", "it", "pt", "pl", "tr", "ru", "nl", "cs", "ar", "zh-cn", "ja", "hu", "ko"]
|
| 99 |
+
return {"total_languages": len(languages), "languages": languages}
|
| 100 |
|
| 101 |
+
@app.post("/tts", tags=["Core Generation"])
|
| 102 |
def generate_tts(
|
| 103 |
+
text: str = Form(..., description="Teks yang akan diubah menjadi suara"),
|
| 104 |
+
speaker: str = Form("Ana Florence", description="Nama karakter suara bawaan dari endpoint /speakers"),
|
| 105 |
+
language: str = Form("id", description="Kode bahasa (contoh: id, en, es)"),
|
| 106 |
+
temperature: float = Form(0.75, description="Kreativitas variasi suara (0.1 - 1.0)"),
|
| 107 |
+
length_penalty: float = Form(1.0, description="Penalti panjang kalimat"),
|
| 108 |
+
repetition_penalty: float = Form(5.0, description="Penalti pengulangan kata berlebih"),
|
| 109 |
username: str = Depends(verify_auth)
|
| 110 |
):
|
| 111 |
+
"""Sintesis teks dasar menggunakan salah satu dari karakter suara bawaan XTTS-v2."""
|
| 112 |
tts = get_xtts_instance()
|
| 113 |
output_path = os.path.join(AUDIO_DIR, f"tts_{uuid.uuid4().hex}.wav")
|
| 114 |
|
|
|
|
| 115 |
available_speakers = tts.speakers
|
| 116 |
if speaker not in available_speakers:
|
| 117 |
+
print(f"Speaker '{speaker}' tidak ditemukan. Menggunakan default: '{available_speakers[0]}'")
|
| 118 |
speaker = available_speakers[0]
|
| 119 |
|
| 120 |
try:
|
| 121 |
+
print(f"Memproses TTS Dasar untuk {len(text)} karakter teks...")
|
|
|
|
| 122 |
tts.tts_to_file(
|
| 123 |
text=text,
|
| 124 |
speaker=speaker,
|
| 125 |
language=language,
|
| 126 |
file_path=output_path,
|
| 127 |
+
split_sentences=True,
|
| 128 |
+
temperature=temperature,
|
| 129 |
+
length_penalty=length_penalty,
|
| 130 |
+
repetition_penalty=repetition_penalty
|
| 131 |
)
|
| 132 |
+
return FileResponse(output_path, media_type="audio/wav", filename="pasblast_tts.wav")
|
|
|
|
| 133 |
except Exception as e:
|
| 134 |
error_trace = traceback.format_exc()
|
| 135 |
+
print(f"FATAL ERROR pada /tts: {error_trace}")
|
| 136 |
+
raise HTTPException(status_code=500, detail=f"Gagal memproses TTS: {str(e)}")
|
| 137 |
|
| 138 |
+
@app.post("/tts_voice_clone", tags=["Core Generation"])
|
| 139 |
def generate_tts_voice_clone(
|
| 140 |
+
text: str = Form(..., description="Teks yang ingin disuarakan oleh hasil kloning"),
|
| 141 |
+
language: str = Form("id", description="Kode bahasa target suara"),
|
| 142 |
+
reference_audio: UploadFile = File(..., description="File WAV sampel suara target (durasi ideal 5-10 detik, bersih dari noise)"),
|
| 143 |
+
temperature: float = Form(0.75, description="Kreativitas variasi suara"),
|
| 144 |
username: str = Depends(verify_auth)
|
| 145 |
):
|
| 146 |
+
"""Zero-shot Voice Cloning: Membuat teks berbunyi persis menyerupai file audio referensi yang diunggah."""
|
| 147 |
tts = get_xtts_instance()
|
| 148 |
|
| 149 |
ref_path = os.path.join(AUDIO_DIR, f"ref_{uuid.uuid4().hex}.wav")
|
|
|
|
| 153 |
output_path = os.path.join(AUDIO_DIR, f"clone_{uuid.uuid4().hex}.wav")
|
| 154 |
|
| 155 |
try:
|
| 156 |
+
print("Memproses Zero-shot Voice Cloning...")
|
| 157 |
tts.tts_to_file(
|
| 158 |
text=text,
|
| 159 |
language=language,
|
| 160 |
speaker_wav=ref_path,
|
| 161 |
file_path=output_path,
|
| 162 |
+
split_sentences=True,
|
| 163 |
+
temperature=temperature
|
| 164 |
)
|
| 165 |
return FileResponse(output_path, media_type="audio/wav", filename="pasblast_cloned.wav")
|
| 166 |
except Exception as e:
|
| 167 |
error_trace = traceback.format_exc()
|
| 168 |
+
print(f"FATAL ERROR pada /tts_voice_clone: {error_trace}")
|
| 169 |
+
raise HTTPException(status_code=500, detail=f"Gagal memproses Voice Cloning: {str(e)}")
|
| 170 |
+
|
| 171 |
+
@app.post("/voice_conversion", tags=["Core Generation"])
|
| 172 |
+
def generate_voice_conversion(
|
| 173 |
+
source_audio: UploadFile = File(..., description="File audio WAV asli berisi ucapan/perkataan seseorang yang ingin diubah suaranya"),
|
| 174 |
+
reference_audio: UploadFile = File(..., description="File audio WAV target berisi sampel karakter suara baru yang ingin ditiru"),
|
| 175 |
+
username: str = Depends(verify_auth)
|
| 176 |
+
):
|
| 177 |
+
"""Voice Conversion (Audio-to-Audio): Mengubah identitas suara pada file audio sumber menjadi karakter suara target tanpa mengubah isi perkataannya."""
|
| 178 |
+
tts = get_xtts_instance()
|
| 179 |
+
|
| 180 |
+
source_path = os.path.join(AUDIO_DIR, f"src_{uuid.uuid4().hex}.wav")
|
| 181 |
+
ref_path = os.path.join(AUDIO_DIR, f"ref_vc_{uuid.uuid4().hex}.wav")
|
| 182 |
+
output_path = os.path.join(AUDIO_DIR, f"vc_{uuid.uuid4().hex}.wav")
|
| 183 |
+
|
| 184 |
+
with open(source_path, "wb") as buffer:
|
| 185 |
+
shutil.copyfileobj(source_audio.file, buffer)
|
| 186 |
+
with open(ref_path, "wb") as buffer:
|
| 187 |
+
shutil.copyfileobj(reference_audio.file, buffer)
|
| 188 |
+
|
| 189 |
+
try:
|
| 190 |
+
if hasattr(tts, "voice_conversion_to_file"):
|
| 191 |
+
print("Memproses Voice Conversion (Audio-to-Audio)...")
|
| 192 |
+
tts.voice_conversion_to_file(
|
| 193 |
+
source_wav=source_path,
|
| 194 |
+
target_wav=ref_path,
|
| 195 |
+
file_path=output_path
|
| 196 |
+
)
|
| 197 |
+
return FileResponse(output_path, media_type="audio/wav", filename="pasblast_converted.wav")
|
| 198 |
+
else:
|
| 199 |
+
raise HTTPException(status_code=400, detail="Model aktif saat ini tidak dikonfigurasi untuk fungsi Voice Conversion bawaan.")
|
| 200 |
+
except Exception as e:
|
| 201 |
+
error_trace = traceback.format_exc()
|
| 202 |
+
print(f"FATAL ERROR pada /voice_conversion: {error_trace}")
|
| 203 |
+
raise HTTPException(status_code=500, detail=f"Gagal memproses Voice Conversion: {str(e)}")
|