HomeTrends
Images
Video
Audio
CanvasMCPAgentNEW
PricingLog inMy creations

AI tools for images, video, and audio

Create content with leading AI models and learn practical AI skills with the IIshki club.

Join the club

Create

  • AI image generation
  • AI video generation
  • Animate a photo with AI
  • Edit a photo with AI
  • AI photoshoot
  • AI text to speech
  • Create music with AI
  • AI models by task

Popular models

  • Seedance 2.5
  • Nano Banana Pro
  • ElevenLabs
  • Veo 3.1
  • Kling 3.0
  • Seedance 2.0
  • GPT Image 2

Platform

  • All AI models
  • AI canvas
  • Compare AI models
  • Alternatives to global AI tools
  • AI chats
  • Blog
  • MCP for ChatGPT and Claude
  • Pricing and tokens

Learning

  • IIshki club
  • Lessons and workshops
  • Personal learning path
  • What the club includes
  • Community ideas
  • Public Offer
  • Privacy Policy
  • Personal Data Consent
  • Publication Consent
  • Cookie Policy
  • AI Content Policy
  • Refund Policy
  • Card Payments
  • Business Details
We acceptELCART
Featured onMossAI Tools
© 2026 Individual Entrepreneur Maksim ShelgunovReg. No. 003-2026-169-3273 of 18 August 2026Support+996 700 254 961Русский
TelegramYouTubeInstagramTikTok
HomeTrendsCreateCreationsProfile
Loading the page…

Kling Motion Control Trends: Crowd Dance, Zombie Hug, Walk Transfer

Published: October 5, 20269 min read

  • kling motion control
  • motion transfer
  • crowd dance trend
  • zombie hug trend
  • transfer walk from video

Kling Motion Control trends — the crowd dance, the zombie hug, music-video walks — all follow one recipe: the photo sets your character's look, and a reference video sets the movement. Upload your photo and a short clip with the motion you want to Kling Motion Control, pick the «By video» orientation, and the model repeats every step, turn and gesture from the clip on you, keeping the background and crowd from the same video. References run 3 to 30 seconds; output is 720p or 1080p. For scenes where the crowd has to be invented from scratch, or where two people interact, Seedance 2.5 is the better fit. Both models run in the IIshki studio.

How Kling Motion Control works: photo + motion

A regular video model guesses the movement from your prompt. Motion Control guesses nothing: it reads the movement from real footage and transfers it onto your image. That is why it works so well for trends — the choreography, the walk and the gestures stay exactly as in the viral clip.

What each file does:

  • Character photo (jpg, png or webp, up to 10 MB, at least 300 px per side) — face, outfit, build. The character can be you, a friend, a mascot or a 3D hero.
  • Reference video (MP4 or MOV, 3–30 seconds, up to 100 MB) — movement, camera angle and surroundings. If the clip has a dance floor, haze and a crowd, they stay in the result.
  • Prompt (optional, up to 2,500 characters) — tweaks to lighting, background and details. The prompt never changes the motion: that always comes from the video.

Audio from the reference is not carried into the result — add the trend's track when editing.

Which model for which trend

Trend Model Why
Crowd dance from an existing clip Kling Motion Control Dance and crowd come from the reference; you replace the dancer
Copy a walk or catwalk from a video Kling Motion Control Steps are copied from real footage, so feet don't drift like in text-to-video
Zombie running at the camera Kling Motion Control One person's motion is applied to a zombie version of your photo
Zombie makeup or a prep frame Nano Banana Pro Changes the look while keeping the face recognizable
A hug between two people, couple + zombie in one shot Seedance 2.5 Holds two characters from photos @Image1 and @Image2, clips up to 30 seconds
A crowd with no suitable reference clip Seedance 2.5 Describe the crowd, club and lighting in the prompt, borrow motion from @Video1

The rule is simple: if you have a clip of one person doing the move, use Motion Control. If you need to invent the scene, or two people interact, use Seedance 2.5.

Where to get a reference video

The reference is 80% of the result. Good sources:

  • The viral clip itself, trimmed to a single shot with no cuts.
  • Your own phone footage: have a friend film you or someone else against a plain background. This is the easiest way to copy a specific walk or dance.
  • Stock footage of a dancer or a running person.

Clip requirements: one main person, body visible fully or at least from the waist up and never blocked; an almost static camera; even lighting; 3 to 30 seconds long. If the clip is longer than 30 seconds or has several cuts, trim it first. Skip clips with recognizable celebrities: the studio's strict moderation rejects them before any tokens are charged.

How to make the trend in Kling Motion Control: 5 steps

1. Prepare the character photo

The key rule: the framing and pose of the photo match the first frame of the video. A full-body dance needs a full-body photo, otherwise the model invents legs that look like someone else's. Hands and feet should be visible, with a little space around the figure. No glare-prone glasses, hats or beauty filters. Upscale a weak or compressed photo first with the upscaler.

2. Upload the files

Open Kling Motion Control: the photo goes into the «Character» slot, the video into «Motion reference». Upload originals, not copies recompressed by a messenger.

3. Pick the orientation

  • By video — up to 30 seconds of motion; camera angle and character position come from the clip. This is the mode for every dance, run and walk trend.
  • By image — up to 10 seconds; the model keeps your photo's composition. Good for gestures and facial expressions when your framing matters more than exact choreography.

4. Pick the quality

Std is 720p, faster and cheaper, for tests. Pro is 1080p with sharper faces and hands, for the final. Cost depends on reference length and quality — the exact figure is shown on the model card before you run it.

5. Add a prompt if needed

You can leave it empty. To change the mood, describe lighting and details, not movement:

Same character, night club, warm spotlights from behind, light haze, the hero is the brightest person in frame, crowd around in half-darkness.

Tips for each trend

Crowd dance

Look for a reference where the dancer stands in the center and is noticeably larger than everyone else, while the crowd moves slower, darker and out of focus. If several people are equally large in frame, the model may pick the wrong person's motion. Use vertical 9:16 — the clip sets the frame, so shoot or crop the reference vertically from the start. For feeds, 5 to 10 seconds works best.

The hero dances in the middle of the dance floor under spotlights, people around are blurred and backlit, pale haze, vertical frame.

If there's no suitable crowd clip, do the dance in Motion Control on any background and build the whole crowd scene in Seedance 2.5 — see below.

Zombie hug

The trend is a tiny story: one partner turns into a zombie and runs, the other lowers a weapon, and it ends in a hug. One generation can't do all of it — split it into 2–3 shots:

  1. Zombie look. In Nano Banana Pro, upload the photo and ask for makeup without blood:

Turn the person in the image into a movie zombie: pale skin, dark circles under the eyes, torn shirt. Face stays recognizable, no blood or wounds, vertical full-body frame.

  1. Running at the camera. This is Motion Control's job: the zombie image goes into «Character», and the reference is a clip of someone running toward the camera and opening their arms. The feet won't slide because the steps come from a real run.
  2. The hug. Two people in contact is where Seedance 2.5 shines: the zombie version as @Image1, the partner as @Image2.

The person from @Image2 lowers a toy blaster, the zombie from @Image1 runs up, and they hug. Cold blue light at the start, warm golden light during the hug, vertical frame, no blood.

Cut the shots together and add music. Cold light at the start and warm light at the end is the main trick that carries the emotion.

Walks, catwalks, music-video shots

To transfer a walk from a video onto your character, you need a clip where the person walks toward or along the camera, feet visible in every frame, with a still or smoothly moving camera. The character photo should be full-body, with shoes, legs not cropped. Use «By video» orientation and pro quality: 1080p shows less foot jitter. Storm-style walks work well — the hero walks against the wind while extras move around; if the extras are already in the reference, they stay.

Common mistakes

Mistake Cause Fix
Feet slide on the floor The reference camera moves a lot, feet are blocked, or the photo is cropped at the knees Use a static-camera clip, a full-body photo with visible feet, pro quality
The face drifts by the end Small face in the photo, glasses, filters, framing that doesn't match the video Evenly lit photo with a large, sharp face and the same framing as the first video frame
The wrong person's motion was transferred Several equally large people in the reference Dancer centered and larger than the crowd; crop around them
«Video too long» error Over 30 seconds, or over 10 in «By image» mode Trim the reference or switch to «By video»
«Video too short» error Under 3 seconds Use a longer segment
Extra fingers, blurry hands Hands in pockets or behind the back in the photo Use a photo with fully visible hands
No sound in the result Reference audio isn't transferred Add the track while editing

When to use Seedance 2.5 instead of Motion Control

Motion Control copies one person's movement from real footage. If there is no suitable clip — the crowd, the club or the zombie street has to be invented — use Seedance 2.5: clips up to 30 seconds at 480p, 720p or 1080p from photos, videos and audio, referenced directly in the prompt as @Image1, @Video1, @Audio1.

Night club, a dense crowd dances in half-darkness, in the center the person from @Image1 is the only sharp, brightly lit figure, dance moves as in @Video1, vertical frame, 10 seconds.

Debug the prompt at 480p and render the final at 720p or 1080p. More ready-made wording is in Seedance 2.5 prompts, and how to recreate a whole movie scene with your character is in the guide to cloning a video. If you're torn between model families, see the Seedance 2.0 vs Kling 3.0 comparison and the best AI tools for photo animation.

What's next

Start with the simplest trend — a dance on your photo from a 5–8 second reference in std. Once you like it, rerun in pro for publishing. For story trends like the zombie hug, assemble the clip from 2–3 shots: Motion Control for exact movement, Seedance 2.5 for scenes with two characters. Use photos of yourself and people close to you, with their consent — other people's faces and celebrities don't belong in trends.

Models in this article

  • Kling Motion Control
  • Seedance 2.5
  • Nano Banana Pro

Frequently asked questions

How does Kling Motion Control work?

You upload two files: a photo of your character and a video with the movement. The photo sets the look, the video sets the motion, camera angle and surroundings. The model repeats the steps, turns, gestures and facial expressions of the person in the clip on your character. A prompt is optional: it can adjust lighting or details, but not the movement.

How do I make the crowd dance trend from my own photo?

Find a clip where one dancer stands large in the center of a dance floor and the crowd around is smaller and blurrier. Upload it to Kling Motion Control as the motion reference, add a full-body photo of yourself as the character, and choose the «By video» orientation. The crowd and club lights come from the clip; the dancer becomes you.

Can I make the zombie trend with Kling Motion Control alone?

Partly. Motion Control transfers the movement of one person, so it is great for the shot where the zombie runs toward the camera with open arms: first make a zombie version of your photo in Nano Banana Pro, then pair it with a running reference. The hug between two people is more reliable in Seedance 2.5 with two photos, @Image1 and @Image2.

How long should the reference video be?

From 3 to 30 seconds, MP4 or MOV, up to 100 MB, with frame sides between 340 and 3850 pixels. In «By video» mode up to 30 seconds of motion is transferred; in «By image» mode up to 10 seconds. Trim long clips first: the cost depends on the reference length.

How do I transfer a walk from a video onto my character?

Shoot or find a clip where a person walks fully in frame, the camera barely moves and the feet are never hidden. The character photo must also be full-body with visible feet. Choose «By video» orientation and pro quality for the final — the steps will match the reference and the feet won't slide.

Every model in this roundup is available in the IIshki studio at reduced club prices. Club members get tokens for generations every month.

Join the club

More articles

  • AI Club Video From Photos: the “Three Meters Above the Sky” TrendHow to make an AI nightclub video from two photos: a ready Seedance 2.5 prompt with @Image1 and @Image2, steps, vertical 9:16 format and common mistakes.October 5, 2026 · 8 min read
  • How to Use Grok Imagine: Image to Video With Sound (2026 Guide)How to use Grok Imagine: turn a photo into video with native sound in version 1.5, make clips from text, pick settings, write prompts and avoid common mistakes.October 2, 2026 · 7 min read
  • Seedance 2.5 Prompts: Formula and 15 Examples With ReferencesHow to write Seedance 2.5 prompts: @Image1, @Video1 and @Audio1 references, shot breakdowns up to 30 seconds, 15 ready-to-use examples and common mistakes.October 1, 2026 · 9 min read

All blog articles →

Related guides

  • Best AI models for video generation
  • Best AI models for image generation
  • Voice a video with AI: voice-over, sound effects and music
  • Animate a photo with AI
  • Edit photos with AI online
  • AI photoshoot
  • AI for marketplaces and product listings
  • Restore old photos with AI
  • Turn a photo into anime with AI
  • AI text to speech
  • Create music and songs with AI
  • Best AI for realistic photos
  • Veo 3.1 vs Kling 3.0 — which should you choose?
  • Kling 3.0 vs Grok Imagine — which should you choose?
  • GPT Image 2 vs Nano Banana Pro — which should you choose?
  • Seedream 5.0 vs GPT Image 2 — which should you choose?
  • Seedream 5.0 vs Nano Banana Pro — which should you choose?
  • Seedream 5.0 vs Grok Image — which should you choose?
  • Seedance 2.0 vs Veo 3.1 — which should you choose?
  • Seedance 2.0 vs Kling 3.0 — which should you choose?
  • Seedance 2.0 vs Grok Imagine — which should you choose?
  • Midjourney alternative — image models in IIshki
  • Stable Diffusion online — the alternative in IIshki
  • Shedevrum alternative — world-class models in IIshki
  • Kandinsky alternative — top models in IIshki
  • Sora 2 alternative — AI video in IIshki
  • ChatGPT in IIshki
  • Claude in IIshki
  • Canvas — wire AI models together on one board

Contents

  1. How Kling Motion Control works: photo + motion
  2. Which model for which trend
  3. Where to get a reference video
  4. How to make the trend in Kling Motion Control: 5 steps
  5. Tips for each trend
  6. Common mistakes
  7. When to use Seedance 2.5 instead of Motion Control
  8. What's next

Models in this article

  • Kling Motion Control
  • Seedance 2.5
  • Nano Banana Pro

Try it in the studio

The models from this article are available in the IIshki studio, paid with tokens.

Open the studio