test

Dockerfile

@@ -1,9 +1,5 @@
 FROM python:3.10-slim
 
-# Install git as root
-USER root
-RUN apt-get update && apt-get install -y git && rm -rf /var/lib/apt/lists/*
-
 RUN useradd -m -u 1000 user
 USER user
 ENV PATH="/home/user/.local/bin:$PATH"
@@ -13,15 +9,6 @@ WORKDIR /app
 COPY --chown=user ./requirements.txt requirements.txt
 RUN pip install --no-cache-dir -r requirements.txt
 
-# Clone MeloTTS and install in editable mode
-RUN git clone https://github.com/myshell-ai/MeloTTS.git /home/user/MeloTTS
-WORKDIR /home/user/MeloTTS
-RUN pip install --no-cache-dir -e .
-WORKDIR /app
-
-# Download unidic dictionary
-RUN python -m unidic download
-
 COPY --chown=user . /app
 CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]
 

app.py

@@ -31,45 +31,20 @@ curl -X POST https://piyazon-tts-piyazon.hf.space/generate-tts \
     -d '{"text": "Hello, world!"}' \
     --output output.wav
 """
-# @app.post("/generate-tts")
-# async def generate_tts(input: TextInput):
-#     try:
-#         # Tokenize input text
-#         inputs = tokenizer(input.text, return_tensors="pt")
-        
-#         # Generate waveform
-#         with torch.no_grad():
-#             waveform = model(**inputs).waveform
-        
-#         # Convert waveform to audio file (WAV format)
-#         waveform = waveform.squeeze().numpy()  # Convert tensor to numpy array
-#         buffer = io.BytesIO()
-#         sf.write(buffer, waveform, samplerate=model.config.sampling_rate, format="WAV")
-#         buffer.seek(0)
-        
-#         # Return audio as streaming response
-#         return StreamingResponse(
-#             buffer,
-#             media_type="audio/wav",
-#             headers={"Content-Disposition": 'attachment; filename="output.wav"'}
-#         )
-#     except Exception as e:
-#         raise HTTPException(status_code=500, detail=f"Error generating audio: {str(e)}")
-
-from melo.api import TTS
-
-# Initialize TTS model for Chinese
-speed = 1.0
-device = 'cpu'
-model_melo = TTS(language='ZH', device=device)
-speaker_ids = model_melo.hps.data.spk2id
-
 @app.post("/generate-tts")
 async def generate_tts(input: TextInput):
     try:
-        # Generate audio to a temporary buffer
+        # Tokenize input text
+        inputs = tokenizer(input.text, return_tensors="pt")
+        
+        # Generate waveform
+        with torch.no_grad():
+            waveform = model(**inputs).waveform
+        
+        # Convert waveform to audio file (WAV format)
+        waveform = waveform.squeeze().numpy()  # Convert tensor to numpy array
         buffer = io.BytesIO()
-        model_melo.tts_to_file(input.text, speaker_ids['ZH'], buffer, speed=speed, format='wav')
+        sf.write(buffer, waveform, samplerate=model.config.sampling_rate, format="WAV")
         buffer.seek(0)
         
         # Return audio as streaming response
@@ -81,6 +56,7 @@ async def generate_tts(input: TextInput):
     except Exception as e:
         raise HTTPException(status_code=500, detail=f"Error generating audio: {str(e)}")
 
+
     
 @app.get("/")
 def greet_json():

requirements.txt

@@ -13,5 +13,4 @@ torchcodec
 flask
 flask-cors
 pydantic
-soundfile
-unidic
+soundfile