The toc toc prompt
A working toc toc prompt has to do five things at once: keep your face the same through every turn, hold a deadpan expression, land the hop on the beat, dress you in the wig and suit without changing who you are, and stop any reference costume from leaking into the result. On Toc Toc Dance AI it is already written and tested for each template. Make the video from one photo.
What the prompt has to solve
- Identity through the turns. The dance goes from front to profile to back and then to the camera again. Each time the face leaves the lens, an AI video model has less of it to work from, and it can slip into a generic face. The wording has to keep the person tied to the photo.
- A deadpan face. The joke only works if the face stays blank while the legs bounce. Models like to add smiles and mouth movement, and a smile in the source photo is easily carried over. The prompt has to say so firmly, and the photo has to help.
- The hop landing on the beat. The trend is a rhythm: one knee at a time, a fist at an invisible door on the beat. A model told only to "dance" will invent its own tempo. The motion has to follow a fixed reference, so the clip stays tied to the song.
- Wig and suit, or your own clothes. In the blond bob and white suit looks the costume replaces your hair and clothes. In the Come As You Are look the opposite is true, and your clothes must survive the hops. Those are two different jobs, so they need two different wordings.
- No leaking from the reference. The motion comes from a real clip of someone else. A careless setup lets that person's face, outfit or room bleed into your video. The reference should lend the movement and nothing else.
Why copied prompts fail
A prompt pasted from a tutorial carries none of the motion reference, input images or rerolls behind the one lucky render it came from, so the face drifts, the hop slips off the beat and costumes leak. Turns can still blur a face and hands vary between renders, even here. The full breakdown is in the blog post on why copied toc toc prompts fail.
What is built in here
Toc Toc Dance AI runs on Seedance 2.0 / 2.5 by ByteDance, and each of its 8 templates is a fixed package: the motion reference, the wording, the room and the settings. They are written once and tested, so every render follows the same shot list of 11–15 seconds and the trend sound is mixed in afterwards. You upload one photo and choose a look, and that is the whole input. The prompt is not shown, because there is nothing for you to edit.
If you want a close-up of how the steps run, read how to make the toc toc video. For the background of the dance, see the trend explained, and for what a render costs, see pricing. A failed or blocked render returns its credits automatically.
The practical difference is who carries the risk. With a pasted prompt every failed attempt is yours, in time and in credits spent elsewhere. Here the choices that usually decide the outcome are already made, and you decide only the look, the format and the photo. A clear, front-facing photo remains the single biggest thing you control.