EzAudio: Enhancing Text-to-Audio Generation with Efficient Diffusion Transformer

🟣 EzAudio is a diffusion-based text-to-audio generation model. Designed for real-world audio applications, EzAudio brings together high-quality audio synthesis with lower computational demands.

🎛 Play with EzAudio for text-to-audio generation, editing, and inpainting: EzAudio

🎮 EzAudio-ControlNet is available: EzAudio-ControlNet

We want to thank Hugging Face Space and Gradio for providing incredible demo platform.

Installation

Clone the repository:

git clone [email protected]:haidog-yaqub/EzAudio.git

Install the dependencies:

cd EzAudio
pip install -r requirements.txt

Download checkponts from: https://huggingface.co/OpenSound/EzAudio

Usage

You can use the model with the following code:

from api.ezaudio import load_models, generate_audio

# model and config paths
config_name = 'ckpts/ezaudio-xl.yml'
ckpt_path = 'ckpts/s3/ezaudio_s3_xl.pt'
vae_path = 'ckpts/vae/1m.pt'
# save_path = 'output/'
device = 'cuda' if torch.cuda.is_available() else 'cpu'

# load model
(autoencoder, unet, tokenizer,
 text_encoder, noise_scheduler, params) = load_models(config_name, ckpt_path,
                                                      vae_path, device)

prompt = "a dog barking in the distance"
sr, audio = generate_audio(prompt, autoencoder, unet, tokenizer, text_encoder, noise_scheduler, params, device)

Todo

Release Gradio Demo along with checkpoints EzAudio Space
Release ControlNet Demo along with checkpoints EzAudio ControlNet Space
Release inference code
Release checkpoints for stage1 and stage2
Release training pipeline and dataset

Reference

If you find the code useful for your research, please consider citing:

@article{hai2024ezaudio,
  title={EzAudio: Enhancing Text-to-Audio Generation with Efficient Diffusion Transformer},
  author={Hai, Jiarui and Xu, Yong and Zhang, Hao and Li, Chenxing and Wang, Helin and Elhilali, Mounya and Yu, Dong},
  journal={arXiv preprint arXiv:2409.10819},
  year={2024}
}

Acknowledgement

Some code are borrowed from or inspired by: U-Vit, Pixel-Art, Huyuan-DiT, and Stable Audio.