YuE Text to Song
Sings your section-labeled lyrics as a track with vocals and accompaniment, using the open YuE models. Required genre, instrument, mood, gender and timbre tags set the style.
Overview
No reference recording goes into this app: YuE builds the voice and the arrangement from your lyrics and the tags you pick. YuE is a family of open lyrics-to-song models co-led by HKUST and M-A-P. Here YuE-s1-7B-anneal-en-cot turns lyrics and tags into a first layer of audio tokens for separate vocal and accompaniment tracks, and YuE-s2-1B-general fills in the remaining layers before decoding.
What you can build
- Demo generators that turn a songwriter's lyric sheet into a sung draft.
- Birthday or wedding songs from verses a user types in.
- Theme songs and jingles for videos.
Inputs
- Lyrics: start each section with a label such as [verse], [chorus], [bridge] or [outro], with a blank line between sections.
- Tags: five required multi-select menus. Genre defaults to Pop (jazz, K-pop and Bossa Nova are among the options), instrument to guitar (Piano, Erhu, Oud), mood to happy (melancholic, Dreamy), gender to female (male, tenor) and timbre to bright vocal (airy vocal, husky vocal). The groups are joined into one space-separated tag line, the form YuE expects.
- Seed (advanced): 0 to 9,999,999, default 42.
Outputs
An MP3 of the final mix, vocal over accompaniment, plus the vocal and the accompaniment as separate MP3 stems. Only the first 10 lyric sections are sung, and in a sheet of 11 sections or fewer the last one is skipped, so a single-section sheet returns no song.
Recommended settings
- Keep each section to a few lines. The YuE authors warn against packing too many words into one.
- Open with [verse] or [chorus]; the authors found [intro] less stable.
- Change the seed for a different take on the same lyrics.
Limitations
The stage-one checkpoint is the English one, with its annealing data mix shifted toward English, so English lyrics are the safe choice. YuE's Chinese and Japanese/Korean checkpoints are not part of this app.
Safety & compliance
YuE, weights included, is released under the Apache License 2.0. The authors encourage creators to use its outputs in their own work, commercial projects included, and to attribute the model as "YuE by HKUST/M-A-P". They recommend labeling published songs as AI-generated and hold users solely responsible for what they generate, including any plagiarism or infringement.
Example prompts
Great starting points for text-to-song-YuE.
API quick start
Run text-to-song-YuE with a single API call.
{
"prompt": "[verse]\nWe’ve been through storms, we’ve …",
"genre": "Pop",
"instrument": "guitar",
"mood": "happy"
}curl -X POST "https://api.wiro.ai/v1/Run/wiro/text-to-song-yue" \
-H "Content-Type: application/json" \
-H "x-api-key: YOUR_WIRO_API_KEY" \
--data-binary @- <<'JSON'
{
"prompt": "[verse]\nWe’ve been through storms, we’ve …",
"genre": "Pop",
"instrument": "guitar",
"mood": "happy"
}
JSONOther models you might like
Explore similar music generation models.



