Cleaned up: switched to runtime model download with huggingface_hub

Dockerfile

@@ -10,39 +10,27 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
     wget \
  && rm -rf /var/lib/apt/lists/*
 
-# Set the working directory
+# Set working directory
 WORKDIR /app
 
-# Copy TTS and install it
-COPY TTS /app/TTS
-WORKDIR /app/TTS
+# Set environment variables
 ENV BLIS_ARCH="generic"
 ENV COQUI_TTS_AGREED=1
-RUN pip install -r requirements.txt --timeout=300
-RUN pip install -e . --timeout=300
 
-# Go back to main app dir
-WORKDIR /app
+# Install Coqui TTS directly
+RUN pip install git+https://github.com/coqui-ai/TTS.git
 
-# Download XTTS model files
-RUN mkdir -p /app/models/xtts_v2
-RUN wget -O /app/models/xtts_v2/config.json https://huggingface.co/coqui/XTTS-v2/resolve/main/config.json?download=true
-RUN wget -O /app/models/xtts_v2/model.pth https://huggingface.co/coqui/XTTS-v2/resolve/main/model.pth?download=true
-RUN wget -O /app/models/xtts_v2/vocab.json https://huggingface.co/coqui/XTTS-v2/resolve/main/vocab.json?download=true
-RUN wget -O /app/models/xtts_v2/dvae.pth https://huggingface.co/coqui/XTTS-v2/resolve/main/dvae.pth?download=true
-RUN wget -O /app/models/xtts_v2/speakers_xtts.pth https://huggingface.co/coqui/XTTS-v2/resolve/main/speakers_xtts.pth?download=true
+# Install other dependencies
+COPY requirements.txt /app/
+RUN pip install -r /app/requirements.txt --timeout=300
 
-# Add speaker reference and other files
+# Copy app files
+COPY local_server_new.py /app/
 COPY audio/speaker_reference.wav /app/audio/speaker_reference.wav
 COPY Web_Page /app/Web_Page
-COPY local_server_new.py /app/
-COPY requirements.txt /app/
-
-# Install app requirements
-RUN pip install -r /app/requirements.txt --timeout=300
 
-# Expose default HF port
+# Expose the port used by Hugging Face Spaces
 EXPOSE 7860
 
-# Run the server directly
+# Start the FastAPI server
 CMD ["python", "-m", "uvicorn", "local_server_new:app", "--host", "0.0.0.0", "--port", "7860"]

local_server_new.py

@@ -8,6 +8,7 @@ import os
 from TTS.api import TTS
 from transformers import AutoTokenizer, AutoModelForSequenceClassification, pipeline
 from langdetect import detect
+from huggingface_hub import hf_hub_download
 
 # Allowlist XttsConfig so torch.load doesn't raise UnpicklingError
 from torch.serialization import add_safe_globals
@@ -32,17 +33,21 @@ app.add_middleware(
     allow_headers=["*"],
 )
 
-# Load TTS model from local files
+# ✅ Dynamically download XTTS v2 model files from Hugging Face
 try:
-    model_dir = "/app/models/xtts_v2"
-    config_path = os.path.join(model_dir, "config.json")
-    # When providing config_path, TTS might expect the directory for model_path
+    print("Downloading XTTS v2 model files from Hugging Face...")
+    model_path = hf_hub_download("coqui/XTTS-v2", "model.pth")
+    config_path = hf_hub_download("coqui/XTTS-v2", "config.json")
+    vocab_path = hf_hub_download("coqui/XTTS-v2", "vocab.json")
+    dvae_path = hf_hub_download("coqui/XTTS-v2", "dvae.pth")
+    speakers_path = hf_hub_download("coqui/XTTS-v2", "speakers_xtts.pth")
+
+    model_dir = os.path.dirname(model_path)
     tts = TTS(model_path=model_dir, config_path=config_path).to("cuda" if torch.cuda.is_available() else "cpu")
-    print("XTTS v2 model loaded successfully from local files.")
+    print("✅ XTTS v2 model loaded successfully.")
 except Exception as e:
-    print(f"Error loading XTTS v2 model from local files: {e}")
-    print("Falling back to loading by model name (license might be required).")
-    tts = TTS("tts_models/multilingual/multi-dataset-xtts_v2").to("cuda" if torch.cuda.is_available() else "cpu")
+    print(f"❌ Failed to load XTTS v2 model: {e}")
+    raise RuntimeError("Failed to initialize TTS model.")
 
 # Load sentiment models
 arabic_model_name = "aubmindlab/bert-base-arabertv02-twitter"
@@ -115,7 +120,7 @@ def text_to_speech(msg: Message):
             text=text,
             file_path=output_filename,
             emotion=emotion,
-            speaker_wav="/app/audio/speaker_reference.wav", # Updated path
+            speaker_wav="/app/audio/speaker_reference.wav",
             language=language
         )
         return {
@@ -129,4 +134,4 @@ def text_to_speech(msg: Message):
 # ✅ Serve the audio file
 @app.get("/audio")
 def get_audio():
-    return FileResponse("output.wav", media_type="audio/wav", filename="output.wav")
+    return FileResponse("output.wav", media_type="audio/wav", filename="output.wav")

requirements.txt

@@ -175,4 +175,5 @@ Werkzeug==3.1.3
 wrapt==1.17.2
 yarl==1.19.0
 zipp==3.21.0
+huggingface_hub>=0.19.0
 # Force rebuild