README
DiffRythmix API Documentation
DiffRythmix is a powerful music generation API that utilizes diffusion models to create high-quality music based on lyrics, audio style references, or text prompts. This API allows developers to generate custom songs with fine-grained control over various parameters.
Table of Contents
- Overview
- Installation & Setup
- API Endpoints
- Testing the API
- Examples
- Response Formats
- Error Handling
Overview
DiffRythmix provides three main functionalities:
- Music Generation: Create original music based on lyrics, style prompts, and other parameters
- LRC Generation from Theme: Generate time-synchronized lyrics based on a theme
- LRC Generation from Lyrics: Convert plain lyrics to time-synchronized LRC format
The API is built using FastAPI and provides both RESTful endpoints and a user-friendly web interface for testing.
Installation & Setup
Prerequisites
- Python 3.10+
- CUDA-compatible GPU (for optimal performance)
- Required packages (see requirements.txt)
Installation
-
Clone the repository:
git clone https://github.com/arslan-sb/RythMix.git git switch backendv1 cd diffrhythm -
Install dependencies:
pip install -r requirements.txt -
Start the API server:
python main.py
The server will start, loading models on startup. This may take a few moments as it prepares the DiffRythmix model, tokenizer, MuQMuLan encoder, and VAE models.
API Endpoints
Generate Music
Generates a music track based on provided lyrics and style parameters.
Endpoint: POST /generate-music
Request Format: multipart/form-data
Parameters:
| Parameter | Type | Required | Description | Default |
|---|---|---|---|---|
| lrc | string | Yes | Lyrics in LRC format with timestamps | - |
| text_prompt | string | No | Text description of the desired music style | "" |
| prompt_type | string | Yes | Type of style prompt to use: "audio" or "text" | - |
| audio_prompt | file | No | Audio file to use as style reference (required if prompt_type is "audio") | null |
| seed | integer | No | Random seed for reproducibility | 42 |
| randomize_seed | boolean | No | Whether to use a random seed | false |
| steps | integer | No | Number of diffusion steps (higher = better quality but slower) | 32 |
| cfg_strength | float | No | Classifier-free guidance strength (higher = more adherence to prompt) | 4.0 |
| file_type | string | No | Output audio format: "wav" or "mp3" | "wav" |
| odeint_method | string | No | ODE integration method: "euler", "midpoint", "rk4", or "implicit_adams" | "euler" |
| music_duration | string | No | Duration of generated music | "95s" |
Response: Audio file stream in the requested format.
Generate LRC from Theme
Generates time-synchronized lyrics based on a theme.
Endpoint: POST /generate-lrc/theme
Request Format: application/json
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
| theme | string | Yes | Theme or topic for the lyrics |
| tags_gen | string | Yes | Style descriptors for the lyrics |
| language | string | Yes | "en" for English or "cn" for Chinese |
Response: JSON with generated LRC format lyrics
Contributors
2 commits
Similar repos
Languages
Python
100.0%
README
DiffRythmix API Documentation
DiffRythmix is a powerful music generation API that utilizes diffusion models to create high-quality music based on lyrics, audio style references, or text prompts. This API allows developers to generate custom songs with fine-grained control over various parameters.
Table of Contents
- Overview
- Installation & Setup
- API Endpoints
- Testing the API
- Examples
- Response Formats
- Error Handling
Overview
DiffRythmix provides three main functionalities:
- Music Generation: Create original music based on lyrics, style prompts, and other parameters
- LRC Generation from Theme: Generate time-synchronized lyrics based on a theme
- LRC Generation from Lyrics: Convert plain lyrics to time-synchronized LRC format
The API is built using FastAPI and provides both RESTful endpoints and a user-friendly web interface for testing.
Installation & Setup
Prerequisites
- Python 3.10+
- CUDA-compatible GPU (for optimal performance)
- Required packages (see requirements.txt)
Installation
-
Clone the repository:
git clone https://github.com/arslan-sb/RythMix.git git switch backendv1 cd diffrhythm -
Install dependencies:
pip install -r requirements.txt -
Start the API server:
python main.py
The server will start, loading models on startup. This may take a few moments as it prepares the DiffRythmix model, tokenizer, MuQMuLan encoder, and VAE models.
API Endpoints
Generate Music
Generates a music track based on provided lyrics and style parameters.
Endpoint: POST /generate-music
Request Format: multipart/form-data
Parameters:
| Parameter | Type | Required | Description | Default |
|---|---|---|---|---|
| lrc | string | Yes | Lyrics in LRC format with timestamps | - |
| text_prompt | string | No | Text description of the desired music style | "" |
| prompt_type | string | Yes | Type of style prompt to use: "audio" or "text" | - |
| audio_prompt | file | No | Audio file to use as style reference (required if prompt_type is "audio") | null |
| seed | integer | No | Random seed for reproducibility | 42 |
| randomize_seed | boolean | No | Whether to use a random seed | false |
| steps | integer | No | Number of diffusion steps (higher = better quality but slower) | 32 |
| cfg_strength | float | No | Classifier-free guidance strength (higher = more adherence to prompt) | 4.0 |
| file_type | string | No | Output audio format: "wav" or "mp3" | "wav" |
| odeint_method | string | No | ODE integration method: "euler", "midpoint", "rk4", or "implicit_adams" | "euler" |
| music_duration | string | No | Duration of generated music | "95s" |
Response: Audio file stream in the requested format.
Generate LRC from Theme
Generates time-synchronized lyrics based on a theme.
Endpoint: POST /generate-lrc/theme
Request Format: application/json
Parameters:
| Parameter | Type | Required | Description |
|---|---|---|---|
| theme | string | Yes | Theme or topic for the lyrics |
| tags_gen | string | Yes | Style descriptors for the lyrics |
| language | string | Yes | "en" for English or "cn" for Chinese |
Response: JSON with generated LRC format lyrics
Similar repos
Contributors
2 commits
Languages
Python
100.0%