How to make a Wipe Me Down dance video: seven steps

To make a Wipe Me Down dance video, you upload one photo or type a description, keep one of the four studio looks selected, choose a resolution and a clip length, press generate, and download the MP4 when the render finishes. The entire job runs in the browser at the studio, there is nothing to install, and a default render (Standard, 720p, 12 seconds) costs 10 credits. Only two of the inputs are really yours to get right: the photo or the description, and the direction you hand the model. Everything else is a setting you can change on the next attempt.

What you need before you start

Three things: a source, an account with credits, and one clear idea of what the dancer should do.

The source is either a photo or a typed description. Photo mode takes one JPG, PNG or WebP file up to 10 MB. Text mode takes a written description of at least 10 characters and no more than 2,000, so you can be as brief as one line or as careful as a paragraph. You cannot supply both at once in a single attempt, and you cannot supply a video clip at all: the model produces a brand-new AI performance from a still image or from words, so nothing here copies or replays anyone's recorded routine.

Creating an account costs nothing, but it comes with no credits: there is no free tier, no signup bonus and no check-in, so a new account starts at 0 and credits are the only thing you buy. One Standard, 720p, 12-second render costs 10 credits, and the cheapest way to try the studio is the $9.90 Single pack, which buys exactly that one clip. There is no free-forever tier and no unlimited plan, so every render comes from a plan or a one-time pack. The pricing page lists all thirteen options with their credit totals, which matters because a long 1080p render costs more than a short 480p one.

The idea is the part people skip. "Dance" is not a direction. Decide before you open the generator whether the subject should hit a two-step bounce, throw a single arm, spin once, or hold a pose at the end, and write that phrase down first.

The seven steps, in order

1. Connect the source. Upload the photo, or type the description if you are working without an image. One file, one piece of text, one subject.
2. Sign in and read the balance. Signing in and any purchase land in the same balance, shown next to the generator. Nothing is granted at signup, so the number you see is the number you bought — confirm it before you spend it, not after.
3. Select one look. The studio ships four: Classic bounce, Hype, Smooth groove and Cute. A look sets the baseline energy, camera feel and lighting, and it is a starting point you can overrule with your own words.
4. Set resolution, length and frame. The default is Standard, 720p, 12 seconds, 10 credits. 480p at the same length is 5 credits, and 1080p at the same length is 60. Frames are 9:16, 16:9 or 1:1, and the studio offers 5, 8, 10 and 12 second clips.
5. Write the direction. Edit the prompt text so it names the subject, the single action, the camera and the light, and says to keep the face, hair and body as supplied. This is the step that separates a usable render from a generic one, and the prompt guide breaks the sentence down piece by piece.
6. Generate, then wait. Most standard renders finish in a couple of minutes; a 1080p render takes longer. Do not resubmit while the job is running, because every submission is charged separately.
7. Download and post. The finished file arrives as an MP4 at the resolution you picked. Keep the source photo and the prompt text, then start the next attempt with exactly one thing changed.

Choosing a photo that survives the render

If you are using photo mode, this decision outweighs every setting on the page. The model keeps the face, hair, body shape and clothing from the image you supply and animates a new performance around them, so the clearer those details are in the source, the more of them survive into the video. A sharp, well-lit photo of one person, taken from the chest up or from head to knees, is the ideal input. The photo guide covers the same ground in more depth, including what to do with older files.

What fails is predictable. A photo where the subject is small in a wide shot gives the model too few pixels to build a face from. A photo where two or three people stand together forces it to decide who dances, and it usually tries to animate all of them. A photo with the subject half in shadow produces a render that inherits the same half-lit face. A blurry or deeply cropped photo produces soft features and occasional limb errors, because the model has to guess the geometry it cannot see.

Lighting and framing

Shoot or choose something where the light falls evenly on the face, ideally from a window or outdoors on a bright but overcast day. Hard side light looks dramatic in a still and confusing in motion, because as the body turns the shadow stays fixed. Frame the subject so there is a little empty space above the head and to the sides; a tight crop that cuts off the top of the head limits how much the model can move the figure without leaving the frame.

One subject beats a group

A single person is not a limitation of the trend, it is a quality setting. One subject means one face to preserve, one body to choreograph and one silhouette to keep stable. If you want two people dancing together, the usual approach is to render two clips from two separate photos and cut them together, rather than asking one render to invent a partner. That also keeps the cost predictable, since each clip is charged on its own spec.

Where your own direction goes

The four looks already contain a workable starting direction, so you can generate without writing anything at all. Classic bounce gives a steady, grounded two-step with a locked-off camera. Hype raises the energy, adds drive to the movement and pushes the camera closer. Smooth groove slows the tempo and softens the lighting. Cute keeps the steps small and bright with a friendlier colour palette. Pick the one that matches the mood of the clip you want, then add a sentence of your own only if you need something the look does not cover, such as a specific arm movement or a final pose.

In text mode there is no photo to anchor identity, so the description carries the whole scene: who is dancing, what they wear, where they are and how the camera behaves. Keep it to a handful of clauses. In photo mode the description is a correction layer rather than a description of the subject, because the image has already settled who the subject is. Both modes accept the same prompt text field, so a sentence that works in one transfers to the other with the subject details removed.

Resolution, length and frames: reading the credit cost

Credit cost follows the spec you choose, not how complicated your prompt is. These are the 12 second figures, which the rest of the table and the longer clips scale from.

| Spec | Clip length | Credits | | ------------- | ----------- | ------- | | 480p | 12s | 5 | | Standard 720p | 12s | 10 | | 1080p | 12s | 60 |

Longer clips scale linearly from those numbers and round up to a whole credit, and the provider accepts anything from 4 to 15 seconds even though the interface offers 5, 8, 10 and 12. The default button is Standard, 720p, 12 seconds, at 10 credits, and that is the figure quoted across this site. Frames are 9:16 for a phone-first feed, 16:9 for a wide screen and 1:1 for a square crop, and the frame does not change the price.

A practical way to spend: test a new idea at 480p for 5 credits, re-render the version you like at 720p, and only move to 1080p once you are confident in both the photo and the words. A 1080p pass costs twelve times a 480p pass, so it is the wrong place to experiment.

What happens while it renders

Renders are private to your account. Nothing is published to a public feed, there is no gallery of other users' clips, and the four looks shown on the site are studio previews rather than a feed of finished jobs. Your clip is the only one you will see.

Most standard renders take a couple of minutes and a 1080p render takes longer, so treat the job as asynchronous and let it finish. If a render fails, the credits return automatically to the balance they came from, so a failure costs you time rather than budget. If you leave the tab while a job is running, come back and check the account rather than pressing generate again.

Downloading and posting

When the job completes you get an MP4 at the resolution you selected, ready for a phone-first feed. Keep the filename sensible and keep the source photo alongside it, because the pair is what makes a second attempt repeatable. Vertical 9:16 clips post without letterboxing to short-form feeds, and square or wide frames suit a feed that shows a grid. The mechanics of re-downloading a clip after closing the browser, and of keeping a record of your renders, are covered in the download guide.

When a render comes back wrong

Failures are usually input problems rather than bad luck, so change the input rather than resubmitting the same pair. Work through this table and change one thing per attempt.

| What you see | Likely cause | First thing to change | | -------------------------------- | -------------------------------------------------------- | ----------------------------------------------------------------- | | Soft or mushy face | Blurry source, or a face that occupies very few pixels | Re-upload a sharp photo where the head fills more of the frame | | Extra limbs or strange arms | The action asks for a pose the photo never shows | Match the action to the reference pose, or pick a simpler gesture | | Two figures where you wanted one | A group photo, or a prompt that mentions a second person | Crop the source to one subject and remove any mention of a crowd | | Motion smears across the clip | Three or four actions requested at once | Rewrite for one dominant action and give it the full duration | | The clip ignores the look | The look was overridden by contradictory prompt text | Empty the prompt text and let the look carry the scene |

If the problem is neither the photo nor the words, it is often the frame. A wide 16:9 frame with a small figure leaves lots of empty space for the model to fill, and 9:16 keeps the subject central by construction. New here? Start with what the trend actually is so you know what you are aiming for, then come back and follow the seven steps.

Frequently asked questions

Can I make a Wipe Me Down dance video from a photo?

Yes, and that is the mode most people use. Upload one JPG, PNG or WebP file up to 10 MB, keep the subject's face, hair and body as they appear in the image, and the model builds a new dance performance around them.

How much does one video cost?

A default render is Standard, 720p, 12 seconds, for 10 credits. 480p at 12 seconds is 5 credits and 1080p at 12 seconds is 60, with longer clips scaling linearly and rounding up. The pricing page shows the credit totals each plan and pack grants.

How long does a render take?

Most standard renders take a couple of minutes, and 1080p renders take longer because there are far more pixels to produce. Wait for the job to finish instead of submitting again, since each submission is a separate charge.

Can I use a video clip as the input?

No. The studio takes a still image or a typed description, and the output is a new AI performance. It does not transfer motion from a reference clip, and it does not swap one real person's face onto another body.

Do failed renders cost me credits?

No. When a render fails, the credits go back automatically to the balance they came from, so the attempt costs time rather than credits.

What should I change if I want a different result?

Change one variable per attempt: the photo, the action, or the look. Changing all three at once leaves you with no way to tell which change worked, and the cheapest way to test a single change is a 5 credit 480p pass.