Follows a supplied lyric sheet closely
Lines you wrote come back sung as written rather than paraphrased or dropped, which is exactly what a songwriter with a finished lyric needs.
YuE
Multimodal Art Projection
The open model that actually sings the words you gave it.
In short
YuE is an Apache 2.0 licensed open-weight song generation model that follows a supplied lyric sheet closely rather than paraphrasing it, and produces full 44.1 kHz stereo output.
YuE specifications
- Output
- Full song with vocals
- Sample rate
- 44.1 kHz stereo
- Architecture
- Autoregressive transformer
- Size
- 7B
- Licence
- Apache 2.0
- Status
- Evaluating
Figures verified 2026-08-10. This field moves quickly — re-check before relying on them.
About YuE
Lyric adherence is a harder problem than it sounds and it is where many song models quietly fail: you supply a finished lyric and the model sings something adjacent to it, dropping lines, repeating others, or reshaping your phrasing to fit a melody it preferred. YuE is notably disciplined here, which makes it the open model of choice for songwriters who arrive with finished words. Combined with Apache 2.0 licensing and full-resolution stereo output, it occupies a genuinely useful position: a self-hostable model that respects authored lyrics and can be deployed commercially without a licensing negotiation. Generation is slower than ACE-Step and the arrangements are more conservative, so this is not the model for exploring directions quickly. It is the one for taking words you have already written and hearing them sung, repeatedly, without paying per attempt.
Strengths
What YuE is good at
Apache 2.0
Apache 2.0 licensing means commercial deployment is an engineering decision rather than a legal negotiation.
Full 44.1 kHz stereo output
Full 44.1 kHz stereo output means the result is release-resolution rather than something you have to upsample and apologise for.
YuE limitations
- Slower than diffusion models
- Heavy GPU requirement at 7B
- Less polished than closed leaders
How it compares
Choosing YuE
What comes back
YuE is an Apache 2.0 licensed open-weight song generation model that follows a supplied lyric sheet closely rather than paraphrasing it, and produces full 44.1 kHz stereo output.
Compare with MusicGenerate
YuE review, pricing and alternatives
Searched as yue, yue ai music, yue model and open source song generator. Lyric adherence is the property that brings songwriters to it specifically.
YuE — questions people ask
Does YuE follow my lyrics exactly?
More literally than most. Lyric adherence is where many song models quietly fail — dropping lines, repeating others, reshaping phrasing — and YuE is notably disciplined about singing what you wrote.
What licence is YuE released under?
Apache 2.0 on the weights, permitting commercial use and modification. As with any open model, that governs the weights and does not itself resolve questions about training data.
What audio quality does YuE output?
Full 44.1 kHz stereo, so the result is release-resolution rather than something you have to upsample.
Is YuE fast?
Slower than ACE-Step, with more conservative arrangements. It is the model for hearing finished words sung repeatedly without paying per attempt, not for exploring directions quickly.
Is YuE free to use?
The weights are free under Apache 2.0, which also permits commercial use — the clearest licensing position of the open song models alongside ACE-Step.
How long does YuE take to generate a song?
Slower than ACE-Step and DiffRhythm. It is the model for hearing finished words sung repeatedly rather than for exploring directions quickly.
Keep exploring
More on YuE and its alternatives
More models
More AI music models and generators
Models and platforms we track, compared on capability and licence.
