Updated Oct 3, 2026
Three layers, one prompt
A rap is really three things stacked together, and the generator builds all three from the line you type:
- Lyrics. A language model writes a short, original verse for two rappers, in English, around your topic.
- Beat. A hip-hop beat is produced to sit under the verse.
- Performance. A Seedance video model renders both of you rapping it, with lip-sync and native audio, on the stage you picked.
You don't edit any of the layers by hand. If you want a different take, change the topic or the stage and make another one; the verse is written from scratch each time.
How to write a topic that raps well
Rap loves detail. A vague topic gets a vague verse; a specific one gets lines your friends will quote back at you.
| Instead of | Try |
|---|---|
| “Work” | “Our boss books meetings at 4:55 on Fridays” |
| “My friend” | “Lena, who has never once been on time” |
| “Summer” | “The road trip where the GPS sent us into a cornfield” |
| “Gym” | “Leg day for two people who hate leg day” |
Add an attitude too: bragging, complaining, celebrating or apologizing. One idea per video works better than five, because a standard verse is only 12 seconds long. You can type up to 200 characters in any language, and the rap will come back in English. Decide who is speaking, too: “I” turns the verse into a confession, “we” turns it into a team anthem.
Topics that sound boring and rap great
- Doing taxes the night before the deadline
- A sourdough starter with a name and a personality
- The office printer that only works for one person
- Assembling flat-pack furniture with zero spare screws
- A group chat that has been renamed 40 times
The trick is contrast. The prouder and bigger the verse sounds, the smaller the subject should be: a parking ticket delivered like a championship anthem is funnier than an actual championship.
No topic at all? Leave the box empty and you get a hype track about the two of you as the greatest duo in the game.
What you control, and what stays fixed
You choose the topic, the two faces, one of six stages, the length (5 to 15 seconds, up to 30 on the Ultra model), the format (vertical, square or widescreen) and the resolution (480p or 720p, with 1080p on Pro and Ultra). A standard 12-second song at 720p is 10 credits and takes about two to three minutes.
Three things stay the same on every render. The verse is in English, the genre is rap over a hip-hop beat, and the voices are generated for your performance rather than copied from any real artist. Want the same topic with a different flavor? Keep the text and swap the stage: a complaint performed in Graffiti Alley feels nothing like the same complaint in the Marble Lobby.