Merge pull request #10 from teddybear082/upgrade_to_pocket_2_0_0

.env.example

@@ -6,6 +6,14 @@ POCKET_TTS_PORT=49112
 POCKET_TTS_LOG_LEVEL=INFO
 POCKET_TTS_STREAM_DEFAULT=true
 
+# Model language (requires pocket-tts>=2.0.0)
+# Options: english, french_24l, german_24l, portuguese, italian, spanish_24l
+# Mutually exclusive with POCKET_TTS_MODEL_PATH
+# POCKET_TTS_LANGUAGE=english
+
+# Enable int8 quantization for lower memory usage and improved speed
+# POCKET_TTS_QUANTIZE=false
+
 # Custom voices directory (mounted to container)
 # POCKET_TTS_VOICES_DIR=./my_custom_voices
 

AGENTS.md

@@ -53,15 +53,21 @@ server.py                    # Entry point, CLI, starts Waitress
 }
 ```
 
-## Configuration (Environment Variables)
-
-| Variable                    | Default   | Purpose            |
-| --------------------------- | --------- | ------------------ |
-| `POCKET_TTS_HOST`           | `0.0.0.0` | Bind address       |
-| `POCKET_TTS_PORT`           | `49112`   | Port               |
-| `POCKET_TTS_VOICES_DIR`     | None      | Custom voices path |
-| `POCKET_TTS_STREAM_DEFAULT` | `true`    | Default streaming  |
-| `POCKET_TTS_LOG_LEVEL`      | `INFO`    | Log verbosity      |
+## CLI Arguments
+
+| Argument               | Env Variable               | Default | Purpose                                         |
+| ---------------------  | --------------------------- | ------- | ------------------                              |
+| `--host`               | `POCKET_TTS_HOST`          | `0.0.0.0` | Bind address                                    |
+| `--port`               | `POCKET_TTS_PORT`          | `49112`   | Port                                            |
+| `--model-path`         | `POCKET_TTS_MODEL_PATH`    | None    | Path to model config file (.yaml) or variant     |
+| `--language`           | `POCKET_TTS_LANGUAGE`      | None    | Model language (english, french_24l, etc.)       |
+| `--quantize`           | `POCKET_TTS_QUANTIZE`      | `false` | Enable int8 quantization for lower memory usage  |
+| `--voices-dir`         | `POCKET_TTS_VOICES_DIR`    | None    | Custom voices directory                          |
+| `--stream`             | `POCKET_TTS_STREAM_DEFAULT`| `false` | Enable streaming by default                      |
+| `--text-preprocess`    | `POCKET_TTS_TEXT_PREPROCESS_DEFAULT` | `false` | Enable text preprocessing by default |
+| `--log-level`          | `POCKET_TTS_LOG_LEVEL`     | `INFO`  | Log verbosity                                    |
+
+**Note**: `--language` and `--model-path` are mutually exclusive. Use `--language` to select a built-in language model (requires pocket-tts>=2.0.0).
 
 ## Voice Resolution Order
 
@@ -79,9 +85,12 @@ server.py                    # Entry point, CLI, starts Waitress
 # Install
 pip install -r requirements.txt
 
-# Run
+# Run with defaults (English model)
 python server.py --log-level DEBUG
 
+# Run with French model and quantization
+python server.py --language french_24l --quantize --log-level DEBUG
+
 # Test
 curl http://localhost:49112/health
 curl -X POST http://localhost:49112/v1/audio/speech \

app/__init__.py

@@ -48,13 +48,20 @@ def create_app(config_overrides: dict = None) -> Flask:
     return app
 
 
-def init_tts_service(model_path: str = None, voices_dir: str = None) -> None:
+def init_tts_service(
+    model_path: str = None,
+    voices_dir: str = None,
+    language: str = None,
+    quantize: bool = False,
+) -> None:
     """
     Initialize the TTS service with model and voices.
 
     Args:
-        model_path: Optional path to model file
+        model_path: Optional path to model config file
         voices_dir: Optional path to voices directory
+        language: Optional language identifier (e.g., english, french_24l)
+        quantize: Whether to apply dynamic int8 quantization
     """
     from app.services.tts import get_tts_service
 
@@ -62,7 +69,7 @@ def init_tts_service(model_path: str = None, voices_dir: str = None) -> None:
     tts = get_tts_service()
 
     # Load model
-    tts.load_model(model_path)
+    tts.load_model(model_path=model_path, language=language, quantize=quantize)
 
     # Set voices directory
     if voices_dir:

app/config.py

@@ -33,6 +33,8 @@ class Config:
 
     # Model settings
     MODEL_PATH = os.environ.get('POCKET_TTS_MODEL_PATH', None)
+    LANGUAGE = os.environ.get('POCKET_TTS_LANGUAGE', None)
+    QUANTIZE = os.environ.get('POCKET_TTS_QUANTIZE', 'false').lower() == 'true'
     DEFAULT_VOICE = os.environ.get(
         'POCKET_TTS_DEFAULT_VOICE', 'hf://kyutai/tts-voices/alba-mackenna/casual.wav'
     )

app/services/tts.py

@@ -61,12 +61,20 @@ class TTSService:
             return str(self.model.device)
         return 'unknown'
 
-    def load_model(self, model_path: str | None = None) -> None:
+    def load_model(
+        self,
+        model_path: str | None = None,
+        language: str | None = None,
+        quantize: bool = False,
+    ) -> None:
         """
         Load the TTS model.
 
         Args:
-            model_path: Optional path to model file or variant name
+            model_path: Optional path to model config file (.yaml)
+            language: Optional language identifier (e.g., english, french_24l).
+                      Incompatible with model_path.
+            quantize: If True, apply dynamic int8 quantization to reduce memory.
         """
         _ensure_pocket_tts()
 
@@ -86,10 +94,13 @@ class TTSService:
         try:
             if effective_path:
                 logger.info(f'Loading model from: {effective_path}')
-                self.model = TTSModel.load_model(config=effective_path)
+                self.model = TTSModel.load_model(config=effective_path, quantize=quantize)
+            elif language:
+                logger.info(f'Loading model with language: {language}')
+                self.model = TTSModel.load_model(language=language, quantize=quantize)
             else:
                 logger.info('Loading default model from HuggingFace...')
-                self.model = TTSModel.load_model()
+                self.model = TTSModel.load_model(quantize=quantize)
 
             self._model_loaded = True
             load_time = time.time() - t0

docker-compose.yml

@@ -25,6 +25,11 @@ services:
       - POCKET_TTS_VOICES_DIR=/app/voices
       - POCKET_TTS_LOG_LEVEL=${POCKET_TTS_LOG_LEVEL:-INFO}
       - POCKET_TTS_STREAM_DEFAULT=${POCKET_TTS_STREAM_DEFAULT:-true}
+      # Model language (e.g., english, french_24l, german_24l, portuguese, italian, spanish_24l)
+      # Mutually exclusive with POCKET_TTS_MODEL_PATH. Requires pocket-tts>=2.0.0.
+      - POCKET_TTS_LANGUAGE=${POCKET_TTS_LANGUAGE:-}
+      # Enable int8 quantization for lower memory usage and improved speed
+      - POCKET_TTS_QUANTIZE=${POCKET_TTS_QUANTIZE:-false}
       # Hugging Face token for voice cloning (optional)
       - HF_TOKEN=${HF_TOKEN:-}
 

pyproject.toml

@@ -4,12 +4,12 @@ license = {text = "MIT"}
 name = "pocket-tts-openai-server"
 readme = "README.md"
 requires-python = ">=3.10"
-version = "2.4.0"
+version = "2.5.1"
 
 dependencies = [
   "flask>=3.0.0",
   "waitress>=3.0.0",
-  "pocket-tts>=1.1.1",
+  "pocket-tts>=2.0.0",
   "torch>=2.0.0,<=2.8.0",
   "torchaudio>=2.0.0,<=2.8.0",
   "scipy>=1.10.0",

requirements.txt

@@ -4,7 +4,7 @@
 # Core dependencies
 flask>=3.0.0
 waitress>=3.0.0
-pocket-tts>=1.1.0
+pocket-tts>=2.0.0
 
 # Audio processing
 torch>=2.0.0,<=2.8.0

run_pocket_tts_server.bat

@@ -68,6 +68,16 @@ set "TEXT_PREPROCESS_ARG=--text-preprocess"
 set /p "INPUT_PREPROCESS=Enable Text Preprocessing? (Y/N) [Y]: "
 if /i "%INPUT_PREPROCESS%"=="N" set "TEXT_PREPROCESS_ARG="
 
+:: 8. Language
+set "LANGUAGE_ARG="
+set /p "INPUT_LANGUAGE=Language (english, french_24l, german_24l, portuguese, italian, spanish_24l - leave blank for default): "
+if not "%INPUT_LANGUAGE%"=="" set "LANGUAGE_ARG=--language %INPUT_LANGUAGE%"
+
+:: 9. Quantization
+set "QUANTIZE_ARG="
+set /p "INPUT_QUANTIZE=Enable int8 Quantization? (Y/N) [N]: "
+if /i "%INPUT_QUANTIZE%"=="Y" set "QUANTIZE_ARG=--quantize"
+
 echo.
 echo ========================================================
 echo Starting Pocket TTS Server...
@@ -77,11 +87,13 @@ if defined MODEL_PATH echo Model: %MODEL_PATH%
 if defined VOICES_DIR echo Voices: %VOICES_DIR%
 if defined STREAM_ARG echo Streaming: Enabled
 if defined TEXT_PREPROCESS_ARG echo Text Preprocessing: Enabled
+if defined LANGUAGE_ARG echo Language: %INPUT_LANGUAGE%
+if defined QUANTIZE_ARG echo Quantization: Enabled
 echo ========================================================
 echo.
 
-:: 8. Run Command
-python server.py --host %HOST% --port %PORT% %MODEL_PATH% %VOICES_DIR_ARG% %STREAM_ARG% %TEXT_PREPROCESS_ARG%
+:: 10. Run Command
+python server.py --host %HOST% --port %PORT% %MODEL_PATH% %VOICES_DIR_ARG% %STREAM_ARG% %TEXT_PREPROCESS_ARG% %LANGUAGE_ARG% %QUANTIZE_ARG%
 
 if %ERRORLEVEL% NEQ 0 (
     echo.

run_pocket_tts_server.sh

@@ -0,0 +1,134 @@
+#!/usr/bin/env bash
+# Pocket TTS Server Launcher (Linux/macOS)
+# This script provides an interactive configuration menu before starting the server.
+
+set -euo pipefail
+
+# Resolve the script directory (handles symlinks)
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+cd "$SCRIPT_DIR"
+
+# Banner
+echo ""
+echo "========================================"
+echo "   Pocket TTS OpenAI Streaming Server"
+echo "========================================"
+echo ""
+
+# --- 0. Hugging Face Token ---
+HF_TOKEN="${HF_TOKEN:-}"
+if [ -z "$HF_TOKEN" ]; then
+    read -r -p "Hugging Face Token (leave blank if already logged in): " INPUT_HF
+    if [ -n "$INPUT_HF" ]; then
+        HF_TOKEN="$INPUT_HF"
+        export HF_TOKEN
+        echo "[INFO] Hugging Face Token set."
+    fi
+fi
+
+# --- 1. Activate Virtual Environment ---
+if [ -d "venv" ]; then
+    echo "[INFO] Activating virtual environment..."
+    source venv/bin/activate
+else
+    echo "[WARNING] 'venv' not found. Using system Python..."
+fi
+
+echo ""
+echo "Configure the server (press Enter to accept defaults):"
+echo ""
+
+# --- 2. Host ---
+read -r -p "Host IP [0.0.0.0]: " INPUT_HOST
+HOST="${INPUT_HOST:-0.0.0.0}"
+
+# --- 3. Port ---
+read -r -p "Port [49112]: " INPUT_PORT
+PORT="${INPUT_PORT:-49112}"
+
+# --- 4. Model Path ---
+echo "Model Config Path (.yaml) or variant name (leave blank for built-in):"
+read -r INPUT_MODEL
+MODEL_ARG=()
+if [ -n "$INPUT_MODEL" ]; then
+    MODEL_ARG+=("--model-path" "$INPUT_MODEL")
+fi
+
+# --- 5. Voices Directory ---
+DEFAULT_VOICES="$SCRIPT_DIR/voices"
+if [ ! -d "$DEFAULT_VOICES" ]; then
+    DEFAULT_VOICES=""
+fi
+PROMPT_VOICES=""
+if [ -n "$DEFAULT_VOICES" ]; then
+    PROMPT_VOICES="$DEFAULT_VOICES"
+fi
+[ -n "$DEFAULT_VOICES" ] && PROMPT_VOICES="$DEFAULT_VOICES"
+read -r -p "Voices Directory [$(if [ -n "$DEFAULT_VOICES" ]; then echo "$DEFAULT_VOICES"; else echo "None"; fi)]: " INPUT_VOICES
+VOICES_ARG=()
+if [ -n "$INPUT_VOICES" ]; then
+    VOICES_ARG+=("--voices-dir" "$INPUT_VOICES")
+fi
+
+# --- 6. Streaming ---
+read -r -p "Enable Streaming? (y/n) [Y]: " INPUT_STREAM
+STREAM_ARG=()
+if [ "${INPUT_STREAM,,}" != "n" ]; then
+    STREAM_ARG+=("--stream")
+fi
+
+# --- 7. Text Preprocessing ---
+read -r -p "Enable Text Preprocessing? (y/n) [Y]: " INPUT_PREPROCESS
+PREPROCESS_ARG=()
+if [ "${INPUT_PREPROCESS,,}" != "n" ]; then
+    PREPROCESS_ARG+=("--text-preprocess")
+fi
+
+# --- 8. Language ---
+echo "Language (english, french_24l, german_24l, portuguese, italian, spanish_24l - leave blank for default):"
+read -r INPUT_LANGUAGE
+LANGUAGE_ARG=()
+if [ -n "$INPUT_LANGUAGE" ]; then
+    LANGUAGE_ARG+=("--language" "$INPUT_LANGUAGE")
+fi
+
+# --- 9. Quantization ---
+read -r -p "Enable int8 Quantization? (y/n) [N]: " INPUT_QUANTIZE
+QUANTIZE_ARG=()
+if [ "${INPUT_QUANTIZE,,}" = "y" ]; then
+    QUANTIZE_ARG+=("--quantize")
+fi
+
+# --- Summary ---
+echo ""
+echo "========================================"
+echo "Starting Pocket TTS Server..."
+echo "  Host         : $HOST"
+echo "  Port         : $PORT"
+[ -n "$INPUT_MODEL" ] && echo "  Model        : $INPUT_MODEL"
+[ -n "$INPUT_VOICES" ] && echo "  Voices       : $INPUT_VOICES"
+echo "  Streaming    : $([ "${INPUT_STREAM,,}" = "n" ] && echo "Disabled" || echo "Enabled")"
+echo "  Preprocessing: $([ "${INPUT_PREPROCESS,,}" = "n" ] && echo "Disabled" || echo "Enabled")"
+[ -n "$INPUT_LANGUAGE" ] && echo "  Language     : $INPUT_LANGUAGE"
+[ "${INPUT_QUANTIZE,,}" = "y" ] && echo "  Quantization : Enabled"
+echo "========================================"
+echo ""
+
+# --- Run ---
+python server.py \
+    --host "$HOST" \
+    --port "$PORT" \
+    "${MODEL_ARG[@]+"${MODEL_ARG[@]}"}" \
+    "${VOICES_ARG[@]+"${VOICES_ARG[@]}"}" \
+    "${STREAM_ARG[@]+"${STREAM_ARG[@]}"}" \
+    "${PREPROCESS_ARG[@]+"${PREPROCESS_ARG[@]}"}" \
+    "${LANGUAGE_ARG[@]+"${LANGUAGE_ARG[@]}"}" \
+    "${QUANTIZE_ARG[@]+"${QUANTIZE_ARG[@]}"}"
+
+EXIT_CODE=$?
+
+if [ $EXIT_CODE -ne 0 ]; then
+    echo ""
+    echo "[ERROR] Server exited with error code $EXIT_CODE."
+fi
+exit $EXIT_CODE

run_pocket_tts_server_exe.bat

@@ -61,6 +61,16 @@ set "TEXT_PREPROCESS_ARG=--text-preprocess"
 set /p "INPUT_PREPROCESS=Enable Text Preprocessing? (Y/N) [Y]: "
 if /i "%INPUT_PREPROCESS%"=="N" set "TEXT_PREPROCESS_ARG="
 
+:: 7. Language
+set "LANGUAGE_ARG="
+set /p "INPUT_LANGUAGE=Language (english, french_24l, german_24l, portuguese, italian, spanish_24l - leave blank for default): "
+if not "%INPUT_LANGUAGE%"=="" set "LANGUAGE_ARG=--language %INPUT_LANGUAGE%"
+
+:: 8. Quantization
+set "QUANTIZE_ARG="
+set /p "INPUT_QUANTIZE=Enable int8 Quantization? (Y/N) [N]: "
+if /i "%INPUT_QUANTIZE%"=="Y" set "QUANTIZE_ARG=--quantize"
+
 echo.
 echo ========================================================
 echo Starting Pocket TTS Server (EXE)...
@@ -70,12 +80,14 @@ if defined MODEL_PATH echo Model: %MODEL_PATH%
 if defined VOICES_DIR (echo Voices: %VOICES_DIR%) else (echo Voices: Default/None)
 if defined STREAM_ARG (echo Streaming: Enabled) else (echo Streaming: Disabled)
 if defined TEXT_PREPROCESS_ARG (echo Text Preprocessing: Enabled) else (echo Text Preprocessing: Disabled)
+if defined LANGUAGE_ARG echo Language: %INPUT_LANGUAGE%
+if defined QUANTIZE_ARG echo Quantization: Enabled
 echo ========================================================
 echo.
 
-:: 7. Run Command
+:: 9. Run Command
 if exist "%~dp0PocketTTS-Server.exe" (
-    "%~dp0PocketTTS-Server.exe" --host %HOST% --port %PORT% %MODEL_PATH% %VOICES_DIR% %STREAM_ARG% %TEXT_PREPROCESS_ARG%
+    "%~dp0PocketTTS-Server.exe" --host %HOST% --port %PORT% %MODEL_PATH% %VOICES_DIR% %STREAM_ARG% %TEXT_PREPROCESS_ARG% %LANGUAGE_ARG% %QUANTIZE_ARG%
 ) else (
     echo [ERROR] PocketTTS-Server.exe not found in the current directory.
     echo Please make sure the executable is located in: %~dp0

server.py

@@ -34,13 +34,15 @@ Examples:
     # Custom port and voices directory
     python server.py --port 8080 --voices-dir ./my_voices
 
-    # Enable streaming by default
-    python server.py --stream
+    # Load French model with quantization
+    python server.py --language french_24l --quantize
 
 Environment Variables:
     POCKET_TTS_HOST         Server host (default: 0.0.0.0)
     POCKET_TTS_PORT         Server port (default: 49112)
-    POCKET_TTS_MODEL_PATH   Path to model file
+    POCKET_TTS_MODEL_PATH   Path to model config file
+    POCKET_TTS_LANGUAGE     Model language (e.g., english, french_24l)
+    POCKET_TTS_QUANTIZE     Enable int8 quantization (default: false)
     POCKET_TTS_VOICES_DIR   Path to voices directory
     POCKET_TTS_STREAM_DEFAULT Enable streaming by default
     POCKET_TTS_TEXT_PREPROCESS_DEFAULT Enable text preprocessing by default
@@ -80,6 +82,19 @@ Environment Variables:
         default=Config.TEXT_PREPROCESS_DEFAULT,
         help='Enable text preprocessing for all requests',
     )
+    parser.add_argument(
+        '--language',
+        type=str,
+        default=Config.LANGUAGE,
+        dest='language',
+        help='Model language (e.g., english, french_24l, german_24l, portuguese, italian, spanish_24l). Incompatible with --model-path.',
+    )
+    parser.add_argument(
+        '--quantize',
+        action='store_true',
+        default=Config.QUANTIZE,
+        help='Apply dynamic int8 quantization to reduce memory usage and improve speed.',
+    )
     parser.add_argument(
         '--log-level',
         type=str,
@@ -106,9 +121,19 @@ def main():
 
     logger = get_logger()
 
+    # Validate mutually exclusive options
+    if args.language and args.model_path:
+        logger.error('--language and --model-path are mutually exclusive. Use one or the other.')
+        sys.exit(1)
+
     # Initialize TTS service
     try:
-        init_tts_service(model_path=args.model_path, voices_dir=args.voices_dir)
+        init_tts_service(
+            model_path=args.model_path,
+            voices_dir=args.voices_dir,
+            language=args.language,
+            quantize=args.quantize,
+        )
     except Exception as e:
         logger.error(f'Failed to initialize TTS service: {e}')
         sys.exit(1)