Spaces:
Sleeping
Sleeping
Add README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,144 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# SingingSDS: Role-Playing Singing Spoken Dialogue System
|
| 2 |
+
|
| 3 |
+
A role-playing singing dialogue system that converts speech input into character-based singing output.
|
| 4 |
+
|
| 5 |
+
## Installation
|
| 6 |
+
|
| 7 |
+
### Requirements
|
| 8 |
+
|
| 9 |
+
- Python 3.11+
|
| 10 |
+
- CUDA (optional, for GPU acceleration)
|
| 11 |
+
|
| 12 |
+
### Install Dependencies
|
| 13 |
+
|
| 14 |
+
#### Option 1: Using Conda (Recommended)
|
| 15 |
+
|
| 16 |
+
```bash
|
| 17 |
+
conda create -n singingsds python=3.11
|
| 18 |
+
|
| 19 |
+
conda activate singingsds
|
| 20 |
+
conda install pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia
|
| 21 |
+
pip install -r requirements.txt
|
| 22 |
+
```
|
| 23 |
+
|
| 24 |
+
#### Option 2: Using pip only
|
| 25 |
+
|
| 26 |
+
```bash
|
| 27 |
+
pip install -r requirements.txt
|
| 28 |
+
```
|
| 29 |
+
|
| 30 |
+
#### Option 3: Using pip with virtual environment
|
| 31 |
+
|
| 32 |
+
```bash
|
| 33 |
+
python -m venv singingsds_env
|
| 34 |
+
|
| 35 |
+
# On Windows:
|
| 36 |
+
singingsds_env\Scripts\activate
|
| 37 |
+
# On macOS/Linux:
|
| 38 |
+
source singingsds_env/bin/activate
|
| 39 |
+
|
| 40 |
+
pip install -r requirements.txt
|
| 41 |
+
```
|
| 42 |
+
|
| 43 |
+
## Usage
|
| 44 |
+
|
| 45 |
+
### Command Line Interface (CLI)
|
| 46 |
+
|
| 47 |
+
#### Example Usage
|
| 48 |
+
|
| 49 |
+
```bash
|
| 50 |
+
python cli.py --query_audio data/query/hello.wav --config_path config/cli/yaoyin_default.yaml --output_audio outputs/yaoyin_hello.wav
|
| 51 |
+
```
|
| 52 |
+
|
| 53 |
+
#### Parameter Description
|
| 54 |
+
|
| 55 |
+
- `--query_audio`: Input audio file path (required)
|
| 56 |
+
- `--config_path`: Configuration file path (default: config/cli/yaoyin_default.yaml)
|
| 57 |
+
- `--output_audio`: Output audio file path (required)
|
| 58 |
+
|
| 59 |
+
|
| 60 |
+
### Web Interface (Gradio)
|
| 61 |
+
|
| 62 |
+
Start the web interface:
|
| 63 |
+
|
| 64 |
+
```bash
|
| 65 |
+
python app.py
|
| 66 |
+
```
|
| 67 |
+
|
| 68 |
+
Then visit the displayed address in your browser to use the graphical interface.
|
| 69 |
+
|
| 70 |
+
## Configuration
|
| 71 |
+
|
| 72 |
+
### Character Configuration
|
| 73 |
+
|
| 74 |
+
The system supports multiple preset characters:
|
| 75 |
+
|
| 76 |
+
- **Yaoyin (ι₯ι³)**: Default timbre is `timbre2`
|
| 77 |
+
- **Limei (δΈ½ζ’
)**: Default timbre is `timbre1`
|
| 78 |
+
|
| 79 |
+
### Model Configuration
|
| 80 |
+
|
| 81 |
+
#### ASR Models
|
| 82 |
+
- `openai/whisper-large-v3-turbo`
|
| 83 |
+
- `openai/whisper-large-v3`
|
| 84 |
+
- `openai/whisper-medium`
|
| 85 |
+
- `sanchit-gandhi/whisper-small-dv`
|
| 86 |
+
- `facebook/wav2vec2-base-960h`
|
| 87 |
+
|
| 88 |
+
#### LLM Models
|
| 89 |
+
- `google/gemma-2-2b`
|
| 90 |
+
- `MiniMaxAI/MiniMax-M1-80k`
|
| 91 |
+
- `meta-llama/Llama-3.2-3B-Instruct`
|
| 92 |
+
|
| 93 |
+
#### SVS Models
|
| 94 |
+
- `espnet/mixdata_svs_visinger2_spkemb_lang_pretrained` (Bilingual)
|
| 95 |
+
- `espnet/aceopencpop_svs_visinger2_40singer_pretrain` (Chinese)
|
| 96 |
+
|
| 97 |
+
## Project Structure
|
| 98 |
+
|
| 99 |
+
```
|
| 100 |
+
SingingSDS/
|
| 101 |
+
βββ cli.py # Command line interface
|
| 102 |
+
βββ interface.py # Gradio interface
|
| 103 |
+
βββ pipeline.py # Core processing pipeline
|
| 104 |
+
βββ app.py # Web application entry
|
| 105 |
+
βββ requirements.txt # Python dependencies
|
| 106 |
+
βββ config/ # Configuration files
|
| 107 |
+
β βββ cli/ # CLI-specific configuration
|
| 108 |
+
β βββ interface/ # Interface-specific configuration
|
| 109 |
+
βββ modules/ # Core modules
|
| 110 |
+
β βββ asr.py # Speech recognition module
|
| 111 |
+
β βββ llm.py # Large language model module
|
| 112 |
+
β βββ melody.py # Melody control module
|
| 113 |
+
β βββ svs/ # Singing voice synthesis modules
|
| 114 |
+
β β βββ base.py # Base SVS class
|
| 115 |
+
β β βββ espnet.py # ESPnet SVS implementation
|
| 116 |
+
β β βββ registry.py # SVS model registry
|
| 117 |
+
β β βββ __init__.py # SVS module initialization
|
| 118 |
+
β βββ utils/ # Utility modules
|
| 119 |
+
β βββ g2p.py # Grapheme-to-phoneme conversion
|
| 120 |
+
β βββ text_normalize.py # Text normalization
|
| 121 |
+
β βββ resources/ # Utility resources
|
| 122 |
+
βββ characters/ # Character definitions
|
| 123 |
+
β βββ base.py # Base character class
|
| 124 |
+
β βββ Limei.py # Limei character definition
|
| 125 |
+
β βββ Yaoyin.py # Yaoyin character definition
|
| 126 |
+
β βββ __init__.py # Character module initialization
|
| 127 |
+
βββ evaluation/ # Evaluation modules
|
| 128 |
+
β βββ svs_eval.py # SVS evaluation metrics
|
| 129 |
+
βββ data/ # Data directory
|
| 130 |
+
β βββ kising/ # Kising dataset
|
| 131 |
+
β βββ touhou/ # Touhou dataset
|
| 132 |
+
βββ resources/ # Project resources
|
| 133 |
+
βββ data_handlers/ # Data handling utilities
|
| 134 |
+
βββ assets/ # Static assets
|
| 135 |
+
βββ tests/ # Test files
|
| 136 |
+
```
|
| 137 |
+
|
| 138 |
+
## Contributing
|
| 139 |
+
|
| 140 |
+
Issues and Pull Requests are welcome!
|
| 141 |
+
|
| 142 |
+
## License
|
| 143 |
+
|
| 144 |
+
|