The Toz Pembe AI trend is a 15-second dance in which three performers in pink Y2K outfits move in sync before a giant boombox, and you can put yourself or your pet in it from one photo. videosi makes the clip from your photos and a reference video, so you write no prompt and learn no steps.
The page already answers the basics. This guide goes further into how to choose photos, why a reference video is used, and what to check before you post.
Key takeaways
- Choose sharp, evenly lit, front-facing photos, waist-up or full-length, with no hats or sunglasses.
- Photo 1 is the right-hand dancer, and empty positions become made-up matching dancers at no extra cost.
- A reference video keeps the choreography and set fixed, so only the dancers change.
- The clip has the trend's sound on it, but platforms may mute it, so a swap to the in-app sound is safer.
- A standard 15-second video is 13 credits, and the Single pack at $6.99 covers a first one.
In this guide
- Which photos work best for a three-person Toz Pembe AI trio?
- Should the right-hand dancer be a pet or a person?
- What happens to a position I leave empty?
- Why does the trend use a reference video instead of a text prompt?
- How does the AI version differ from the official choreography?
- Can I post the video with the original sound?
- Should I label the video as AI-generated?
- Which format should I pick: 9:16, 1:1 or 16:9?
- How many credits does the first video cost, and what about a retry?
- Is the trend early enough to be worth posting?
Which photos work best for a three-person Toz Pembe AI trio?
Pick three photos that look like they were taken in the same kind of light, even if they were taken years apart. The dancers stay full-body in a wide, fixed shot, so each face ends up small. The model has very few pixels to work with, and a photo that is sharp, front-facing and evenly lit gives it the most to hold on to.
Framing matters as much as sharpness. A waist-up or full-length photo lets the model see build and hair length, which a tight face crop cannot. Avoid hats and sunglasses, because the dancers turn and step and anything covering the face makes the likeness drift. If one of your three photos is much darker or blurrier than the others, that dancer will be the one who looks off, so swap it for a better one before you spend credits.
Order counts too. Photo 1 is the dancer on the right, Photo 2 the left and Photo 3 the middle. If one person matters more to the post, such as the birthday friend or the family pet, put them in slot 1. You can see all three slots on the Toz Pembe AI page.
Should the right-hand dancer be a pet or a person?
Both work, and the difference is in how the model reads the photo. A pet stays that animal, with its own markings, and dances upright on its hind legs in the pink outfit. A person stays a human dancer with no cat ears, whiskers or tail, even though the reference clip is full of cats.
Pets reward a clear photo with the colours visible, facing the camera. A cat with a distinctive coat, such as tortoiseshell patches or a white bib, stays recognisable because the pattern is the identity. A plain black cat in a dim photo has little for the model to hold, so the result looks generic. People have the opposite problem: faces are what viewers check first, and a wide shot makes faces small, so a person needs the sharpest photo of the two.
You can mix them. A trio of one person and two pets, or three pets of different species, is allowed, because each photo is used for exactly one dancer.
What happens to a position I leave empty?
An empty position becomes a made-up dancer of the same kind as your first photo. If Photo 1 is a pet, the other two become other animals of that species with different colouring. If Photo 1 is a person, they become different ordinary adults. They wear the same pink outfits the reference clip gives those positions.
That makes a one-photo video a complete trio, not a solo with gaps. It is also the cheapest way to test a photo: adding more photos does not change the credit cost, so you only fill the other slots when you actually have people or pets you want in the clip.
Why does the trend use a reference video instead of a text prompt?
Because the trend is a piece of choreography, and words are a poor way to describe three people moving in sync. A model given only a typed prompt tends to lose the timing between dancers, drift in camera framing and change the set from shot to shot. A 15-second reference clip fixes the moves, the beat, the giant boombox, the promenade and the wide camera, and the prompt only has to say who stands where.
This is why the clip you get matches the one people have seen: low sway, arms open, side steps with hip pops, arms overhead, then a closing group pose. The model swaps the dancers and leaves everything else as it was. It runs on ByteDance's Seedance, with the tier you choose on the page.
It also explains why you do not need to write anything. The long prompt that drives the swap is built in and is not published, since a copied prompt without the reference clip does not reproduce the effect.
How does the AI version differ from the official choreography?
The official dance clips come from the group themselves. "Toz Pembe" is a single the Turkish girl group Manifest released on 3 July 2026, a date a Turkish news report on the release also gives. Fans copied that choreography at home, which is where the trend started.
The AI version is something else. It is a template based on the three-cat clip posted in October, with your photos in place of the cats. You do not learn the steps, you do not film, and you do not need two friends in the same room. The trade-off is that you get one fixed routine, not your own interpretation, and the moves are what the reference clip shows rather than the group's original steps.
If you want to dance it yourself, the fan clips are the better guide. If you want your cat or your group chat on a promenade in front of a boombox, the AI version is the only practical route. videosi is independent and has no connection with Manifest or its label.
Can I post the video with the original sound?
The clip arrives with the trend's sound already on it, so it plays correctly the moment you download it. Whether a platform keeps that sound is a separate question. The song belongs to its rights holders, and a platform may mute or limit a post that uses it from an uploaded file.
The safer routine is simple. Upload the clip, and if the app offers "Toz Pembe" in its own licensed sound library, swap to that version. Posting inside the app with its sound library is the way the platform expects a song to be used, and it avoids a muted clip after you have shared it with friends.
Should I label the video as AI-generated?
Yes. Platforms increasingly ask for an AI-generated label, and it takes one tap. A label also saves an awkward comment thread: the clip shows cats or friends dancing in sync on a seaside promenade, and viewers will wonder how you filmed it.
Use only photos of adults who agreed, or your own pets. The tool does not accept photos of children, celebrities or public figures, and a render with such a photo is blocked and refunded. If you are putting friends in the trio, tell them before you post, because a dancing version of someone is still their likeness.
Which format should I pick: 9:16, 1:1 or 16:9?
Match the format to where the clip will live. 9:16 fits TikTok, Reels and Shorts, and fills a phone screen. 1:1 suits a feed post or a group chat. 16:9 is for a wide cut, such as a slide, a website banner or a YouTube upload.
The set is built wide, so a vertical crop leaves the dancers larger in frame and the boombox partly cut, while 16:9 shows the whole promenade and makes each face smaller. If likeness is your worry, choose 9:16 or 1:1 so faces take up more pixels. If the set is the point, go wide.
How many credits does the first video cost, and what about a retry?
Prices come from the plans on the site. A standard 15-second video, using the Standard model at 720p, costs 13 credits. An account's first video is capped at the price of a 12-second one on the same tier, so the Single pack of 10 credits at $6.99 covers a first Toz Pembe video. Later videos pay by the second.
If you plan to make several versions, the Trio Pack is 30 credits for $17.99, the Trend Pass is $9.99 a month for 30 credits, and the Duo Pack is 50 credits for $29. Credits never expire. There is no free generation, and a failed, timed-out or blocked render returns its credits.
Spend sensibly on retries. If a result looks wrong, the photo is the likely cause, not bad luck. Change the photo first, then regenerate.
Is the trend early enough to be worth posting?
It is early as an AI format. The song charted in Turkey over the summer, and a Soundcharts YouTube chart for Turkey listed "Toz Pembe" by manifest when we checked it. The three-cat dance appeared on 4 October 2026, and template pages started listing it that month. On 29 October 2026 our trend radar flagged it as budding, not peaking.
Early formats reward whoever posts a recognisable version first, and fewer people have posted a pet or a friend group version than a plain copy of the cats. A personal twist, such as your own dog in the right-hand slot, is the part nobody else has.
