The mandem in disguise AI trend is a 7-second TikTok joke in which one photo of a group flips to a lads version and then to masked mandem in black puffers, on a clip from Skepta's Disguise. You can make it from one photo of you or your friends on videosi, with the sound already on the video.
This guide goes deeper than the trend page FAQ: what the words mean, where the sound comes from, how the three parts are built, which photo to choose, and who to make it with.
Key takeaways
- The trend is a three-state photo reveal with two hard cuts over 7 seconds, not a smooth transformation.
- The sound is a clip from Skepta's 2009 track Disguise, featuring JME and Jammer.
- Use one to five adults in a sharp, front-facing photo for the best result.
- Captions are not rendered, so add your own text at about 0 s, 3 s and 4 s.
- Use photos of friends who are happy to be in the clip, never to mock anyone, and label the video as AI.
In this guide
- What does "mandem" actually mean in this trend?
- What is the Skepta Disguise song behind the sound?
- How is the three-part structure built?
- Why does the trend use hard cuts instead of a smooth morph?
- How do you pick the right group photo?
- How does the AI version get made from one photo?
- CapCut filter or videosi: which should you use?
- How do you add the captions?
- Do you need to label the video as AI?
- Who should you make it with?
- What does it cost to make one?
What does "mandem" actually mean in this trend?
"Mandem" is British slang for a group of lads, a crew of mates. In this trend a still photo of a group flips to a lads version of the same people, and then to masked lads in black puffer jackets, hence "mandem in disguise".
The usual captions spell the joke out: "They're not girls / They're guys / They're mandem in disguise". The final state plays on the word "disguise": black puffer jackets and balaclavas or hoods with only the eyes showing. It is the punchline, which is why the third beat is held longest on screen.
What is the Skepta Disguise song behind the sound?
The sound is a short clip from Disguise, a 2009 track by Skepta featuring JME and Jammer. The clip uses the short part with the "mandem in disguise" hook. On TikTok it usually appears as an original sound rather than under the song title, so searching for the title there can miss it.
Skepta is a grime artist, and grime is a genre that emerged in London in the early 2000s, built on fast, syncopated beats and MC vocals. Skepta and his brother JME are credited with founding the Boy Better Know crew. That context explains why "mandem" is the natural word: the whole scene is built around crews.
How is the three-part structure built?
The format is one still photo shown in three states over 7 seconds, with two cuts and no camera move.
- 0 to 3 s, the photo. The group exactly as they are.
- 3 to 4 s, the lads. The same photo, same poses and places, but everyone is a lad in tank tops and jeans.
- 4 to 7 s, the mandem. The same poses again, now in black puffers and balaclavas with only the eyes showing, held to the end.
The timing follows the spoken line on the sound. The first state sets up the false premise, the second delivers the reveal in one second, and the third is the punchline you let sit. If you film or edit this yourself, keep the middle beat short. A long second beat kills the rhythm.
Why does the trend use hard cuts instead of a smooth morph?
The reveal only works as a cut. A morph would show people gradually changing, which turns a joke about a sudden switch into a transformation effect. A hard cut keeps the poses frozen and lets the viewer's brain do the comparison: same spots, same smiles, different people.
This is also the part generators find hardest. Models like to blend between states, so the structure has to be pinned down firmly. On videosi's mandem in disguise page the structure is built in: a frozen photo, two hard cuts and no camera move. One known limit is that the cut can land a few tenths of a second off the original sound, so if the timing matters to you, check it before posting.
How do you pick the right group photo?
The photo decides the result, because it is the only source for faces, build, pose and background in all three states. A few rules that work:
- Use one to five adults. Beyond five, faces start to blur or merge, especially in the masked state.
- Pick a sharp, well-lit photo with everyone facing the camera and fully inside the frame. Dim club shots and heavy filters make the lads version drift.
- Avoid big hats, costumes and sunglasses on everyone. They confuse the first change.
- Group selfies and mirror photos work well, since people already pose close together.
You can also use a solo selfie, a couple or an all-male group. A person who already is a man stays himself in the middle state and then gets the masked look. The lads version is the model's guess at how each person would look as a man, so it will not match anyone's expectations exactly.
How does the AI version get made from one photo?
On videosi you upload one photo, pick 9:16, 1:1 or 16:9, and press Generate. There is nothing to write. The tool uses ByteDance's Seedance models in a photo plus motion reference mode: your photo supplies each person's face, build, pose and background, and a short reference clip supplies the structure of one frozen picture with two hard cuts.
The instruction behind it is built in and not shown. Its job is to replace the group in the reference clip with the people in your photo and to hold the three states over 7 seconds. The result is a new video, not an edit of anyone's original post. It takes a few minutes, and if you close the page it is saved in My creations.
Common failure modes are worth knowing. If the replace instruction is weak, the model keeps the people from the reference and only borrows your background. If the lads version is not tied to each person's own face shape, age and skin tone, it drifts toward generic men. And a model can turn the cuts into a slow morph. The mandem in disguise AI page handles all three, so you do not tune them yourself.
CapCut filter or videosi: which should you use?
On TikTok most posts of this trend use a CapCut photo filter, shared as a template link, that is aimed at a group shot. It is quick and free to try, and it is how the format spread. The trade-off is that you depend on that template and on how well it reads your particular photo.
videosi works differently. It does not use that filter; a short clip of the format guides only the timing of the cuts. It generates a new 7-second video from your photo, with the Disguise sound already on it and a no-watermark MP4 as output. Choose CapCut if you want the quick template route and are happy with its result. Choose videosi if you want a generated video built around your own photo, a choice of video tier, and the audio lined up with the three states.
How do you add the captions?
The captions are deliberately not rendered into the video. Burned-in AI text often comes out garbled, so videosi leaves it off and you add it yourself in TikTok or CapCut. Place your own lines at the cuts, for example one from about 0 s, a second from about 3 s and a third from about 4 s.
Keep the text large, centred in the safe area above the TikTok interface, and use the same font for all three so the cuts feel clean. Do not cover the faces, since the faces are the whole joke.
Do you need to label the video as AI?
Yes, label it. The video is AI-generated, and platforms increasingly ask creators to disclose that. Use the platform's own AI-generated content label when you upload, and say so in the caption if you like. It costs nothing and avoids confusion, especially because the lads version shows real people with faces that they did not actually have.
Only upload photos of adults you have permission to use. If a friend is in the photo, ask them before you post. This takes ten seconds and is the difference between a joke and a problem.
Who should you make it with?
Make it with friends who are happy to be in the clip and would find it funny. Only upload photos you have permission to use, ask anyone in the photo before you post, and do not use the trend to mock or target anyone. If you are unsure whether your group would like it, skip it; there are plenty of other photo formats on videosi.
What does it cost to make one?
A standard 7-second video (Standard model, 720p) is 10 credits, and the higher tiers cost more. A Single pack is 10 credits for $6.90, so it covers one standard video. There is no free tier. A failed, timed-out or blocked render returns its credits automatically, so a bad run does not cost you.
Because each try takes several minutes and uses credits, spend the effort on the photo first.
