AI music production, stable-audio-tools local deployment experience

Introduce:

Stable Audio Tools is an open source audio generation library. We can use any prompt words to generate various music, including various audios. Next we are going to build a free AI audio generator locally.

Deployment:

git stable-audio-tools hub.

git clone https://github.com/Stability-AI/stable-audio-tools.git 
cd stable-audio-tools

Create a virtual environment:

pip install -U virtualenv
python -m virtualenv stable-audio-venv
source stable-audio-venv/bin/activate

Install dependency packages:

pip install .

Model download:

Model download requires downloading the stable-audio-open-1.0 pre-training model and the google/t5-base text processing model.

Enter the stable-audio-tools warehouse and create the folders stabilityai/stable-audio-open-1.0 and t5-base to store the stable-audio-open-1.0 model and t5-base model respectively.

cd stable-audio-tools
mkdir -p stabilityai/stable-audio-open-1.0
mkdir t5-base

Enter the website https://huggingface.co/stabilityai/stable-audio-open-1.0 and download.

Enter the website https://huggingface.co/google-t5/t5-base/tree/main and download.

Run:

The operation of Stable-Audio-Open-1.0 is very simple. After activating the virtual environment, run the stable-audio-tools/run_gradio.py file (you can pass — model-config /path/to/model/config — ckpt-path /path /to/wrapped/ckpt loads the model_config.json and model.safetensors files from a custom path).

source stable-audio-venv/bin/activate
python3 run_gradio.py --model-config stabilityai/stable-audio-open-1.0/model_config.json --ckpt-path stabilityai/stable-audio-open-1.0/model.safetensors

Open the web page: http://127.0.0.1:7860 (Music prompt words can be found on the official website https://stableaudio.com/user-guide/prompt-structure)