laitimes

A more controllable AI music generation tool – Sonauto

author:The mountain monster Atu

Yesterday, the AI music generation tool Sonauto.ai released their version 1.0, which may not be as great as Udio and Suno, but Sonauto has more editing features that make AI music generation more controllable.

A more controllable AI music generation tool – Sonauto

Peculiarity:

Sonauto can convert cues, lyrics, and melodies into full songs in any style. It is a latent diffusion model rather than a language model. A typical music generation model converts music into tokens by training vector quantization variational autoencoders, such as the Description Audio Codec (https://github.com/descriptinc/descript-audio-codec), and then trains a large language model on these tokens to generate music. Instead, Sonauto removed the tokenization and replaced it with the normal variational autoencoder bottleneck. and train a diffusion transformer (similar to Sora) in a nice, normally distributed latent space. Sonauto is the first audio diffusion model to generate coherent lyrics.

Sonauto has a tempo control mode where you can upload your own percussion lines or set up BPM. It is also possible to generate appropriate variations for uploaded or previously generated songs (for example, you can even sing a one-minute voice memo and upload it).

How to use:

用邮箱登录sonauto.i,点制作一首歌Make a song.

A more controllable AI music generation tool – Sonauto

Create a song title

A more controllable AI music generation tool – Sonauto

You can select an artist style, enter the artist's name in the input box, and a search result will appear, such as "John Denver", and an automatic prompt will appear below, such as "John Denver's voice tells the story clearly." Country and classic rock elements blend perfectly in the arrangement...", or type in a cue yourself, such as "a John Denver song about home".

A more controllable AI music generation tool – Sonauto

Enter the lyrics

A more controllable AI music generation tool – Sonauto

Advanced settings can be set to control the tempo intensity, the number of beats, or upload reference songs directly.

A more controllable AI music generation tool – Sonauto

After the point is generated, 3 songs will be generated for 1 minute and 35 seconds

A more controllable AI music generation tool – Sonauto

Finally, you can also edit the generated songs, one is to split. Split the song into individual instruments, vocals, bass, and more.

A more controllable AI music generation tool – Sonauto

The second is to expand the song.

A more controllable AI music generation tool – Sonauto

The third is that the rhythm control can be reset.

Let's listen to two songs and see the effect

Loading...

Loading...

Overall, Sonauto is also an excellent AI music generation model. Unfortunately, it currently does not recognize Chinese lyrics and cannot generate songs with correct Chinese pronunciation.

Recommended in the past

Udio released a beta version, and Sora, the AI music industry, exploded

Earn 800 a day, how Xiaobai uses AI music to make money, gameplay and Suno.AI detailed explanation

Tiangong Skymusic started internal testing and electrocuted the first domestic AI music tool

Stablity AI发布Stable Audio 2.0,挑战Suno AI

Read on