Tiangong Music

Kunlun Tech's AI music generation product built on the TianGong large model can achieve end-to-end generation of musical instruments, human voices and melodies, and supports lyric paragraph control, a variety of music styles and dialect singing, according to the official statement.

AI Music Tools#AI Music#Song Generation#Songwriting

Updated: 2026-10-26

What is Tiangong Music?

SkyMusic, an AI music generation product launched by Kunlun Tech based on the TianGong large model, adopts a Sora-like DiT architecture, according to public data, enabling end-to-end integrated generation of musical instruments, human vocals and melodies. The homepage of the product is divided into two entrances: AI Songwriting and Discover Great Songs, which lowers the threshold for songwriting for ordinary users.

Who is it for?

Music enthusiasts with zero foundation: First write lyrics, then use AI to compose and sing, creating private works such as birthday songs and commemorative songs.
Short video creators: Quickly generate mood-matching background music according to themes, avoiding repeated screening in copyright music libraries.
Music professionals: Use reference audio to test different music styles and singing plans, and use the finished products as a reference for arrangement and singing.
Dialect content creators: Create regional songs such as Cantonese and Chengdu dialect for localized dissemination.
Student clubs: Quickly get finished demo tracks with vocals for events, to be used for rehearsals and promotions.

Key features

Audio Generation

The official states that it combines Large-scale Transformer and Diffusion Transformer to achieve end-to-end generation of instruments, vocals, melodies, volume and musical notes.

Lyric Control

The official states that it supports lyric paragraph control, which can reflect the emotional differences between verses and choruses, as well as between the intro and verses, facilitating the expression of delicate emotions.

Style Adaptation

Public information shows it supports a variety of music styles such as rap, folk, funk, ancient style, and electronic music, and you can also upload reference audio to specify the desired style.

Vocal Techniques

The official states that it can learn and express singing techniques such as vibrato, opera, chanting, male-female duets and automatic harmonization, making the singing more in line with the emotional tone of the work.

Dialect Singing

According to public information, the product supports the generation of songs in dialects such as Cantonese and Chengdu dialect, which can be used for regional content creation.

Work Release

The product features a dedicated section called "Discover Great Songs", where other users can listen to, like and share the works, facilitating a cycle of creation and sharing.

How to use Tiangong Music

First clarify the theme and mood before writing lyrics.
the clearer the segmentation of the lyrics, the more distinct the emotional differences between the sections.
Choose reference audio as material that you have the right to use, and avoid directly imitating protected mature works.
The copyright ownership and commercial licensing terms for AI-generated songs are subject to change with version updates.
you must review the current agreement before commercial use.
When using dialects or specific vocal tones, listen to a short clip first, confirm the pronunciation is natural before generating the full track, and reduce rework.
It is recommended that the generated songs undergo additional mixing processing in professional tools, and the rights boundaries should be confirmed before public use.
In addition, the copyright ownership and commercial authorization of AI music are unavoidable issues: most platforms stipulate that tracks generated by free accounts are only for personal non-commercial use, and the authorization scope corresponding to the subscription level must be confirmed before official release or commercial projects.