Spaces:
Build error
Build error
| """ | |
| Bark TTS microservice for Ebonex AI — Hugging Face Spaces deployment. | |
| Bark (by Suno) generates highly expressive, emotional audio from text. | |
| It supports laughter, breathing, whispering, and dramatic storytelling — | |
| perfect for the Ebonex Kids Mode. | |
| Arabic voice presets: v2/ar_speaker_0 through v2/ar_speaker_9 | |
| English voice presets: v2/en_speaker_0 through v2/en_speaker_9 | |
| Recommended for storytelling: | |
| Arabic: v2/ar_speaker_9 (warm, expressive narrator) | |
| English: v2/en_speaker_6 (clear, warm storyteller) | |
| """ | |
| import base64 | |
| import io | |
| import os | |
| import numpy as np | |
| import soundfile as sf | |
| from fastapi import FastAPI, HTTPException | |
| from fastapi.middleware.cors import CORSMiddleware | |
| from pydantic import BaseModel | |
| from bark import SAMPLE_RATE, generate_audio, preload_models | |
| # Pre-load models on startup (takes ~30s, only once) | |
| # Set SUNO_USE_SMALL_MODELS=True for faster but lower-quality output on CPU | |
| use_small = os.environ.get("SUNO_USE_SMALL_MODELS", "True").lower() == "true" | |
| preload_models( | |
| text_use_small=use_small, | |
| coarse_use_small=use_small, | |
| fine_use_small=use_small, | |
| ) | |
| app = FastAPI( | |
| title="Bark TTS for Ebonex", | |
| description="Storytelling TTS powered by Suno Bark", | |
| version="1.0.0", | |
| ) | |
| app.add_middleware( | |
| CORSMiddleware, | |
| allow_origins=["*"], | |
| allow_methods=["POST", "GET"], | |
| allow_headers=["*"], | |
| ) | |
| class TtsRequest(BaseModel): | |
| text: str | |
| voice_preset: str = "v2/ar_speaker_9" | |
| def root(): | |
| return {"status": "ok", "service": "Bark TTS for Ebonex"} | |
| def health(): | |
| return {"status": "ok"} | |
| def text_to_speech(req: TtsRequest): | |
| if not req.text or not req.text.strip(): | |
| raise HTTPException(status_code=400, detail="text is required") | |
| # Bark works best with texts under 400 characters. | |
| # For longer texts, split and concatenate (handled by TtsService on Node side). | |
| text = req.text.strip()[:400] | |
| try: | |
| audio_array = generate_audio(text, history_prompt=req.voice_preset) | |
| except Exception as e: | |
| raise HTTPException(status_code=500, detail=f"Bark generation failed: {str(e)}") | |
| # Convert numpy float32 array → WAV bytes → base64 string | |
| buffer = io.BytesIO() | |
| sf.write(buffer, audio_array, SAMPLE_RATE, format="WAV", subtype="PCM_16") | |
| buffer.seek(0) | |
| audio_b64 = base64.b64encode(buffer.read()).decode("utf-8") | |
| return { | |
| "audio": audio_b64, | |
| "sample_rate": SAMPLE_RATE, | |
| "format": "wav", | |
| "voice_preset": req.voice_preset, | |
| } | |