The shadow boxing AI trend is an 8-second clip in which one photo of an ordinary person, usually a grandparent, turns into a jab, slip, roll and uppercut combo thrown at the air. On videosi you make it from one standing photo with nothing to write, and the video comes out silent. This guide covers what the basic pages do not: how it differs from the Ish game, why the joke works, what the moves mean and what to check before you post.
Key takeaways
- Shadowboxing AI is a one-photo video, not the two-person pointing game also called the Ish game.
- A full-body standing photo gives the best likeness, and likeness is never guaranteed.
- Other tools often give slow, floaty punches because they lack a motion reference.
- The video has no sound, so choose your own music, and label the post as AI.
- A standard 8-second video (Standard model, 720p) costs 7 credits, and failed renders return credits.
In this guide
- Is shadowboxing AI the same as the shadow boxing trend or Ish game?
- Why is the shadowboxing meme almost always a grandparent?
- What do jab, jab, slip, roll and uppercut actually mean?
- How do I shadow box properly in real life?
- What song is used in the shadow boxing TikTok?
- Why do ChatGPT, Sora, Veo, Kling or CapCut give slow, floaty punches or a younger face?
- Why does the face not look like the person, and which photo should I pick?
- Can I use a pet photo for shadowboxing AI?
- What should I check before posting a shadowboxing AI video?
- How long does it take and how much does it cost?
Is shadowboxing AI the same as the shadow boxing trend or Ish game?
No. The shadow boxing game that TikTok explainers cover is a two-person pointing game, and the shadowboxing AI trend is a one-photo video. Know Your Meme describes the older game, also called the Ish game or Pointing Game, as a social game that went viral in late 2019 and was revived on TikTok in 2023. Two people face each other and take turns pointing left, right, up or down, and the goal is to make the other person look the same way. You can read the entry on Know Your Meme.
The AI trend has no rules, no opponent and no pointing. One ordinary photo becomes an 8-second clip of that person boxing the air. So if you searched for "shadow boxing rules" or "game explained", you want the pointing game, which needs two real people. If you searched for "jab slip uppercut" with a grandma in the caption, you want the shadowboxing AI page.
Why is the shadowboxing meme almost always a grandparent?
Because the joke is the contrast. A grandparent in a cardigan in the living room is the least likely person to square up to nobody, and the clip plays it with a completely straight face. Captions such as "Alright grandma relax" do the rest.
The spread is also practical. The trend travels as a CapCut template called "Shadowboxing filter", and the people who share templates mostly post a family photo with a funny caption. We have not confirmed who started it, so we do not name anyone. The earliest remakes we found on TikTok were posted on 12 and 14 September 2026, and almost all came from template-sharing accounts.
The template itself is easy to check. Its CapCut page shows an 11-second template titled "Shadowboxing filter" with 226.77K uses when we looked on 23 October 2026. Treat it as a snapshot.
What do jab, jab, slip, roll and uppercut actually mean?
The combo in the clip is a list of boxing basics, and the Wikipedia article on boxing styles and technique describes each one.
- Jab: a quick, straight punch thrown with the lead hand from the guard position. Wikipedia says its power comes from the legs, not the arm.
- Slip: moving the head slightly offline of an incoming punch, often by leaning and twisting the upper body.
- Roll (bob and weave): bobbing moves the head sideways and beneath an incoming punch, then the boxer weaves back up on the outside or inside of the arm.
- Uppercut: a vertical, rising punch. The rear version is thrown with the rear hand.
In the videosi clip these land in a fixed order over 8 seconds: a bounce into stance, a duck, a reset, two jabs and a straight, then the slip, roll and uppercut, and finally a high guard. Nobody throws a real punch at anyone, and the video has no sound.
How do I shadow box properly in real life?
Start with a coach, because a video of a grandmother is not a lesson. The Wikipedia entry on shadowboxing says it is a combat sports exercise where you throw punches at the air as if there were an opponent, that it mainly refines punching technique, and that trainers often use it as a warm-up to loosen up and keep rhythm.
That is useful guidance on what it is for, but the details of stance and form are better learned in person. General safety points for any beginner: warm up first, don't snap your elbows straight at full extension, leave clear space, and stop if anything hurts. People with joint, heart or balance problems, and that includes many of the grandparents in the trend, should ask a doctor before trying a workout. Please do not copy the clip's speed at home, because it is a rendered video and not a training plan.
What song is used in the shadow boxing TikTok?
We could not verify one. There is no single confirmed song for the shadowboxing AI trend: the remakes we watched use different sounds, many posted by template-sharing accounts, and we have not found one original audio that all of them share. The CapCut template page we opened also lists no music. Songs that happen to be titled "Shadowboxing" by other artists are unrelated, and we have not confirmed any link to them.
The videosi video is a silent 8-second MP4, so the choice is yours. Pick any sound from the TikTok or Instagram library when you post, or add music in your editor.
Why do ChatGPT, Sora, Veo, Kling or CapCut give slow, floaty punches or a younger face?
Mostly because they are guessing the motion. If you type "grandma shadowboxing, jab jab slip roll uppercut" into a general video tool, the model has no clear idea of real-speed punches. You tend to get slow, floaty arm waves, no real duck, and a head that stays still. And once an older person starts boxing, video models like to make them younger and more muscular, which is the opposite of the joke.
The fixes are structural. The videosi page uses ByteDance's Seedance model with a five-part prompt (identity lock, motion reference, scene, timeline and rules) and a short motion reference clip of a real boxer that supplies the footwork and timing. The identity lock names age and build first so the person stays who they are. We do not publish the full wording. If you use another tool, look for one that accepts a reference video, and expect several tries, each costing time and credits.
CapCut is a different case. Its template is a filter applied inside CapCut with its own rules, and templates like it change or disappear often. videosi is independent and not affiliated with CapCut or TikTok.
Why does the face not look like the person, and which photo should I pick?
The photo is the usual cause. A face crop, sunglasses, a hat or a seated pose gives the model too little to hold on to. Use one standing photo from head to shoes, in even light, with the face clear and nothing covering it. The video is filmed full body from a still camera, so the model needs the whole body as well as the face.
Likeness is never guaranteed. It drifts most in the two moments when the head turns side-on or ducks, and fast punches are where detail fails first, so a blurred hand or an extra finger can appear in the jab and the uppercut. If the first clip is off, change the photo instead of repeating the same one.
Can I use a pet photo for shadowboxing AI?
You can try, but it is not what the page is tuned for. Dog and cat shadowboxing clips are part of the wider trend, but the motion on the shadowboxing AI page is built for a person on two legs, with a human boxer's footwork and fists. A pet photo can be uploaded, but a usable result is not guaranteed.
If you want an animal in a human role, a standing pet photographed head to feet gives the model the most to work with.
What should I check before posting a shadowboxing AI video?
Three things. First, label it as AI. Platforms ask for it when a clip shows a realistic person, and an older relative boxing in a living room looks real enough that the label matters. Second, ask the person in the photo before you post, and show them the result. The trend works best when the grandparent is in on the joke. videosi does not accept photos of celebrities, public figures or minors, and users must be 18 or older.
Third, do not present the clip as a real fight or real footage. It is a joke video of someone throwing punches at the air, so avoid captions that suggest a real confrontation or a real skill. We also use no footage of any real fighter. Some captions link the combo to a famous fighter's old training clip, and we have not confirmed that link. Photos and results are deleted within 30 days, so download the MP4 you want to keep.
How long does it take and how much does it cost?
Rendering takes several minutes, and if you close the page the result is saved in My creations. You upload one JPG, PNG, WEBP or HEIC photo up to 10 MB, pick 9:16, 1:1 or 16:9, and press Generate. There is no prompt to write.
A standard 8-second video (Standard model, 720p) costs 7 credits. There is no free generation and no free trial. Credit packs start at $6.99 for 10 credits and never expire, and subscriptions start at $29.90 a month. Better models and higher resolutions cost more. A failed, timed-out or blocked render returns its credits. To try it, open the shadowboxing AI trend page.
