Function 1   Generate music from prompts.

Emotion and mood

Please select the music emotion and mood. No more than 2 items.


Style

Please select a music Style.

Function 2   Accompaniment Generation

Please select the mood and style prompts on the page.

Upload music files with .mid .midi extension.

Music will be generated automatically after uploading

The MIDI music file is required to contain only the vocal track, or name the vocal melody track "melody".

Technical features

Our generative model is high-quality, efficient, low-cost, controllable, interactive and easy to call.

We provide service interface that only requires a few lines of code to request and get results.

1. Efficient and low-cost
With our efficient algorithms, the AI model only relies on a single 4090 card and can create a piece of midi in less than 1s !

Most of the working time is spent on audio mixing (midi-to-audio).
2. Controllable and interactive
General controllable generation technology: "Labeled data is all you need".

We has developed AI modification and filling, but it requires the support of DAW and editing platform. Looking for business cooperation.
3. high-quality and easy to use
Objective evaluation such as model perplexity and note distribution of different instruments are currently ranked first.

It only takes a few lines of code to request and get the generated results.

Contact Us

Advisor: haonancheng@cuc.edu.cn

Researcher: zouyiiy@sina.com

Welcome to use our model interface and technical services!

Warning

This website is only used for AI algorithm testing! Not for commercial use!

Currently, the amount of training data is small, far from the amount of commercial platforms.

Therefore, there will be similar samples, which can be completely solved after expanding the model parameters and data volume.