Template, your own clip or a motion model

Invisible Moped AI: The Crazy Frog Chase From a Photo or Your Own Moves

Invisible moped AI is the Crazy Frog AI trend: a person crouches in mid-air on a motorcycle that is not there, races along an overpass, looks back at a cop on a white hover-bike, backflips and escapes. Use the photo template here for the trend's timing in 15 seconds, or film your own crouch and move it to a road when you want your real movement.

  • One or two photos · 15 seconds
  • 9:16 or 16:9 · trend soundtrack
  • People or pets · five models
  • From 36 credits

制作 Crazy Frog AI 视频

Who chases you?

一张骑手照片,警察由模型虚构。

Photo 1 · Rider (on the invisible moped)

300–6000 px wide and high, up to 30 MB. One person or pet per photo, facing the camera in even light, no sunglasses. Waist-up or full-body photos give the crouch and the flip a real body to work with. Adults only — use your own photo, or one from someone who said yes.

Where the chase happens

The overpass at dusk is the trend’s look. Pick another place, or upload an empty photo of your street, office or school — the chase moves there.

分辨率
画幅

16:9 is the trend clip’s wide frame, with the cop and the road beside you. 9:16 follows the rider for TikTok, Reels and Shorts, so the flip stays in frame. Same price.

时长

480p · 9:16 · 15 s MP4 · the trend sound on the video

消耗 130 积分

照片和角色参考图以私有方式保存,生成失败自动退款。 隐私政策

追逐示例

The original overpass, cut for cut
Invisible moped AI at a glanceOn this page
What you add

One rider photo, or two photos for rider and cop. People or pets; JPG, PNG or WebP up to 30 MB, 300–6000 pixels per side. No prompt.

What you get

A 15-second MP4 with the chase soundtrack, 9:16 (set by default) or 16:9, on the overpass, Neon City, Supermarket or your own background photo.

Price

From 36 credits; 130 with the default Seedance 2.5 at 480p. One or two photos cost the same.

Failed task

Credits return automatically. A finished take you dislike is not refunded.

WatermarkBefore your first purchase, downloads carry a moving watermark. Any purchase unlocks the clean originals, including videos already made in your account.

There is no frog in the video. The name comes from the song: creators lay Crazy Frog’s 2005 “Axel F” under a person who rides a motorcycle that does not exist. The AI version spread on TikTok in October 2026; the overpass chase posted by @wdmowski on 12 October is the one most remakes copy.

Photo, your own clip or a motion model: which invisible moped AI route fits

People make the same chase in three ways. They differ in where the movement comes from. The invisible moped AI generator above takes a photo and supplies the motion. Your own crouch clip keeps your body. A general invisible motorcycle AI model makes you bring a reference and a prompt.

RouteYou uploadMovement comes fromOriginal cuts and timingTwo peopleSound
Chase template (this page)1–2 photos, optional place photoThe built-in chaseFollowed beat by beat; the model can still vary a poseYes: rider and cop slots, SwapTrend soundtrack included

Your crouch clip in the

Genjutsu AI editor

Your MP4/MOV (5–300 s) and 1–9 place or person images

Your own body and cameraOnly as close as you film itFilm both people, then change the placeYour clip’s own audio
Higgsfield GenjutsuA source video (4–30 s), up to 30 imagesThe source video (Motion Transfer or Object Swap)Follows the source you supplyNot statedNot stated
Kling Motion ControlA character image and a motion videoThe motion videoFollows the referenceOne primary subjectReference audio not used
Seedance 2.5Images and a reference videoThe reference, interpretedFraming kept, details reinterpretedNot statedGenerates its own audio

Checked 19 October 2026 on each product’s own page; Kling 4.0, released this month, also takes video references, so check Kling’s documentation for its limits. The table compares inputs, not quality: we have not run the same photo through every tool. If you only care about looking like the trend, the template is the shortest route. If you want people to see you actually holding the pose, film it.

The chase second by second

The remakes that read as “the invisible moped trend” keep seven beats in the same order. Use this map to judge a result, to film your own take, or to time a prompt for another model. Seconds are approximate for a 15-second clip.

SecondsWhat happensWhat to check
0–3

Low front view: the rider crouches in mid-air, fists on handlebars that aren’t there, gliding along the overpass.

Nothing under the rider; feet off the road.
3–5

The rider comes toward the camera; a hover-bike appears far behind.

The face matches your photo before the first cut.
5–7Close-up of the cop on the white hover-bike.

In solo mode this is a stranger; in duo mode, Photo 2. It should not be the rider’s face.

7–9The rider looks back over the shoulder as the cop closes in.The profile still looks like the rider; hands stay closed.
9–11Backflip, then a landing with both hands thrown up.Limbs attached, one body, the flip lands where the music hits.
11–13Low wide shot as the rider pulls away.No real bike appears in the getaway.
13–15An aerial pull-back over the road.The setting you chose is still there.

The sound is Crazy Frog’s “Axel F” (2005). That character goes back to a 1997 recording of a teenager imitating a moped engine, which animator Erik Wernquist turned into the 2003 “Annoying Thing” animation. If you add the sound yourself to a clip you filmed, slide it in the editor until your landing falls on a strong beat.

Film your own crouch: invisible moped from a video of yourself

A still photo gives the model your face and clothes, but not how you move. If you want your own crouch, lean and flip in the video, record it and change everything around you instead. This is the route for a real squat or a dancer’s flip, and for how to do the invisible moped trend without a stranger’s body.

Make the invisible moped chase from your own crouch clip

  1. Film the crouch. Prop the phone at hip height and record yourself holding a deep crouch with both fists forward as if gripping handlebars. Lean into imaginary turns, look back over your shoulder near the middle, then throw both hands up. Keep the whole body in frame.
  2. Upload it to the editor. Open the Genjutsu AI editor and upload the MP4 or MOV (5–300 seconds, up to 50 MB). Slide to the window you want, up to 15 seconds on most models.
  3. Choose Background and add the road. Pick the Background change and add a photo of an empty overpass or street as Image 1. In the instruction, ask for a fast-moving road and a cop on a white hover-bike behind you. The place is changed; your body and camera stay.
  4. Check, then add the sound. Watch for a real bike appearing under you and for the chaser drifting in and out. The editor keeps your clip’s own audio, so add the Axel F sound in TikTok, Reels or Shorts when you post.

Filming tips that match the beats: shoot from below hip height so the camera looks slightly up at you, as in the trend; hold the crouch for the first five seconds; look back between seconds 7 and 9; do the jump or flip around second 9 only if it is safe, on a soft surface; and stop moving forward so the editor can supply the speed. A plain wall behind you makes the new place easier to read.

What the editor will and won’t do: it changes the place and keeps your body, outline and camera. A chaser you describe in the instruction is a request, not a guaranteed character with its own close-up. If you want the cop’s close-up and the exact cut order, the photo template is the better fit. To put a different person on your movement instead, use motion transfer.

Rider and cop from two separate selfies

You don’t need a photo of the two of you together. Switch the template to duo, add the rider as Photo 1 and the cop as Photo 2, and use Swap if you picked the wrong way round. Each slot should show one person (or pet) with a visible face. The rider owns the opening, the look back and the flip; the cop owns the 5–7 second close-up and appears behind the rider in the wide shots. For a version you film yourselves, record both people in one clip and change the place in the editor; duo swap recasts a clip that already has two people.

Four takes that keep the beats

Each example below follows the same chase and soundtrack as the template, in a different setting.

The original overpass, cut for cut

Squat, cop close-up, look back, flip at about 9.6 seconds, hands-up landing and the aerial pull-away land on the same beats as the trend clip.

Same beats, Halloween road

The chase keeps its order while the setting turns into a graveyard road and the cop wears a pumpkin helmet.

Overpass and chocolate factory

The location switches between the overpass and a candy factory across cuts; the flip and landing stay on the beat.

Indoors on a mall floor

An indoor setting with a police scooter; the squat, flip and low getaway follow the same timing.

When the motion goes wrong

SymptomLikely causeWhat to do
The flip happens early or not at allA prompt-only model invents its own timing

Use a motion reference (the template, or your own clip) and name the second of the flip in the prompt.

The cop appears in the rider’s shots, or the close-up is missingShot order not anchored to a reference

Describe the cop close-up as its own shot at 5–7 seconds, or use the template.

A real motorcycle or scooter shows upThe model completes the riding pose with a vehicle

Say “no motorcycle, scooter, seat or wheels at any time”; check the getaway shot first.

Hands open, legs change positionA weak pose in the reference or source photo

Film a firmer crouch with closed fists; use a photo where the hands are visible.

Rider and cop swap faces at a cutRoles not assigned, or similar photos

One person per photo; keep Photo 1 as rider and Photo 2 as cop; use Swap rather than re-uploading.

The chosen place turns back into an overpassThe chase motion carries road shapes into later shots

Use a clear, empty place photo and check the final aerial shot before posting.

A new completed take is charged again; failed tasks are refunded. Higher resolution makes the picture sharper but does not fix timing or anatomy.

Credits for one 15-second take, two routes

The same 15 seconds costs a different amount depending on the route. The template includes a private still made from your photo and the chase finishing; your own clip pays only for the edit of your footage.

Model · resolutionChase templateYour own clip (editor)
Seedance 2.5 · 480p130 credits ≈$4.83117 credits ≈$4.35
Seedance 2.5 · 720p275 credits ≈$10.22262 credits ≈$9.74
Seedance 2.5 · 1080p478 credits ≈$17.77464 credits ≈$17.25
Seedance 2.0 · 480p95 credits ≈$3.5381 credits ≈$3.01
Seedance 2.0 · 720p187 credits ≈$6.95174 credits ≈$6.47
Seedance 2.0 · 1080p449 credits ≈$16.69436 credits ≈$16.21
MiniMax H3 · 720p129 credits ≈$4.80116 credits ≈$4.31
Seedance 2.0 Fast · 480p62 credits ≈$2.3148 credits ≈$1.78
Seedance 2.0 Fast · 720p117 credits ≈$4.35104 credits ≈$3.87
Seedance 2.0 Mini · 480p36 credits ≈$1.3423 credits ≈$0.86
Seedance 2.0 Mini · 720p62 credits ≈$2.3148 credits ≈$1.78

Dollar figures are the credits’ value at the lowest monthly rate ($29 for 780 credits), not a per-video checkout price. New accounts receive 3 credits, valid for 7 days. One full clip starts at 36 credits. The $5 Trial pack contains 24 credits. The pack alone does not cover a full clip. If all 3 sign-up credits are still valid and unused, the combined 27 credits still do not cover the minimum-price clip. The cheapest own-clip edit is 23 credits. See plans and credit packs.

Invisible moped AI prompt for other motion models

If you use Higgsfield Genjutsu, Kling or Seedance with a reference video, ask it to change only the rider and keep everything else. Use a clip you are allowed to use, such as your own crouch take. Adjust the wording to the tool’s own reference syntax.

Replace only the rider
Reference video: the invisible moped chase clip. Reference image: the new rider.
Replace only the rider with the person in the reference image. Keep their face, hair, glasses and clothes in every shot.
Keep everything else from the reference video: the crouched riding pose with both hands on handlebars that are not there, the overpass, the cop and the white hover-bike, every camera angle and cut, the look back, the backflip and the hands-up landing at the same moments, and the getaway.
There is no motorcycle, scooter, seat or wheels under the rider at any time. Feet stay off the ground while riding.
One rider only. Do not change the cop's face. No text, captions, logos or watermarks.

Add the 15-second beat map above if the tool accepts timed instructions.

Two people: rider and cop
Reference video: the invisible moped chase clip. Image 1: the rider. Image 2: the cop.
Image 1 replaces the crouching rider in every rider shot, including the look back and the backflip.
Image 2 replaces the cop on the white hover-bike, including the close-up of the cop.
Never give the rider the cop's face or the cop the rider's face. Never merge the two people.
Keep the reference's cuts, timing, camera moves and the invisible bike: nothing under the rider.

Kling Motion Control states it supports one primary subject, so a two-person chase may need a tool that accepts several images.

Rights: the reference clip, the song and other people’s faces

  • Someone else’s video. A viral invisible moped clip belongs to its creator. Feeding it to a motion tool to publish your own version can still use their work; film your own or use a template.
  • The song. “Axel F” by Crazy Frog is a commercial recording. Add it from the platform’s sound library for ordinary posts; an ad or a brand account needs a licence, and some platforms restrict music on business accounts.
  • Other people’s faces. Ask before you upload a friend’s or coworker’s photo, even for a private chat. Our terms allow only adults who agreed to the edit; tell viewers the video is AI-generated where the platform asks you to label it.

This is general information, not legal advice.

最近更新: 2026-10-23

Invisible moped AI questions

Is this the Higgsfield Genjutsu people mention for the invisible moped trend?

No. Genjutsu AI is an independent site and is not Higgsfield. Higgsfield Genjutsu is a separate product with Motion Transfer and Object Swap modes. This page has a ready-made chase template that needs one or two photos, and a video editor where you can upload your own clip and change the place or the person.

Do I need to download the original TikTok to use as a motion reference?

Not for the template on this page. The chase motion, cuts and soundtrack are already built in, so you only add photos. You would need a source clip only for a general motion-transfer model, and then it should be footage you have the right to use, such as your own crouch video.

Why does the flip come too early, or the cop show up in the wrong shot?

Prompt-only video models invent their own timing, so the backflip and the cop close-up drift. A motion reference fixes the order of shots: the template here follows the chase beat by beat. If you use another model, give it the clip as a video reference and say which beat happens when, using the second-by-second map on this page.

In a two-person chase the faces swap between shots. Is that the photos or the roles?

Usually the roles. When both photos are sent without a clear assignment, a model can give the cop’s face to the rider at a cut. Here Photo 1 is always the rider and Photo 2 always the cop, and Swap exchanges them. Use one person per photo, each with a visible face, and avoid two near-identical photos.

A real motorcycle appears under the rider. What should I change?

The model is filling in the pose with the object it expects. In your own prompt, state that there is no motorcycle, scooter, seat or wheels at any moment and that the feet stay off the ground. A clearly crouched reference helps more than extra adjectives. On the template, check the low getaway shot, where it is most likely to appear.

Can I feed someone else’s viral invisible moped video into a motion-transfer tool?

The video belongs to whoever made it, even if you replace the person and the background. Using it to make your own post can infringe their rights or the platform’s rules. Film your own crouch, use a template, or ask the creator. This is general information, not legal advice.

Which Crazy Frog sound is it, and is it in the video I download?

The trend uses Crazy Frog’s 2005 recording of “Axel F”. A template download from this page includes the chase soundtrack. A clip made from your own footage keeps your own audio, so add the sound inside the app you post to. Before your first purchase, downloads carry a moving watermark. Any purchase unlocks the clean originals, including videos already made in your account.

Can CapCut alone make the invisible moped effect?

Not the generated chase. The CapCut “Crazy Frog” template arranges your own footage with the song; it does not create the overpass, the hover-bike cop or the flip. You can film a crouch and edit it in CapCut, but the moving road and the chaser have to come from a video model or a lot of compositing.

用任意 AI 助手核实

“How do I make the invisible moped AI video, and should I use a photo template, my own clip or a motion-transfer model?”

保留这条镜头,改换其中的世界。

选择片段,添加参考素材,几分钟后即可看到编辑结果。生成前会先显示报价。

打开 Genjutsu AI 编辑器