How to Make an AI Dance Video From a Photo

To make an AI dance video from a photo, you give an AI model two things: a clear full-body photo of the person (or character) and a short reference video of the dance. A motion control model such as Kling V3 Pro Motion Control reads the moves from the video and applies them to your photo, so the person in the picture performs the routine with the same timing. Of the three methods covered here, it is the only one that follows real choreography.

A dance template is the quickest route: one click applies a preset trend dance at a flat price. Image-to-video with a dance prompt needs no clip at all, because the model improvises movement from text. Neither holds to a specific routine.

The 3 ways to make an AI dance video, compared

Most searches for an AI dance generator end at one of these three. Our take: if the dance matters (a trend, a routine, a music video moment), use motion control. If you just want a fun clip of your cat doing a belly dance, use a template. Image-to-video with a dance prompt is the method most people try first and the one most likely to disappoint, because a prompt like "dancing to hip hop" gives the model no actual choreography to follow.

Method 1: motion control from a reference dance clip (best)

Motion control, also called motion transfer, is a video-to-video mode. It reads the body positions of the dancer in your clip and retargets them onto the person in your photo. The Kling motion control guide covers every setting in depth. This section sticks to what matters for dance.

Step by step on Oakgen

1. Film or pick the reference clip. See the checklist below. Trim it to the section you want, because the reference length sets the charge (you pay per second of it) and roughly the output length. The output can come out shorter, especially with fast motion.
2. Open Kling V3 Pro Motion Control. If you do not have a clip yet, the form offers 11 preset motion clips under the upload field, which are useful for a first test.
3. Motion Reference: upload your trimmed dance clip.
4. Character: upload your photo.
5. Character Orientation: keep Match Video. It allows up to 30 seconds and makes the character turn and move the way the dancer does, which is what you want for dance. Match Image caps clips at 10 seconds and keeps the pose direction from your photo.
6. Description (optional): describe the scene or outfit changes and leave the moves to the clip. For example: "dancing in a neon-lit parking garage at night, wet floor reflections".
7. Advanced: turn Keep Original Sound off if the clip has music you cannot publish (see the music section). If the face must stay recognizable, add a facial element: a front-facing face photo plus side views.
8. Generate a 5-second test first, check hands and face on the fastest move, then run the full clip.

How to pick the reference dance clip

The reference clip decides more about quality than the photo does. Every rule below comes from Kling's own Motion Control guide or the model's input limits.

• One dancer only. If several people are in frame, Kling uses whoever takes up the most of the frame.
• Whole body and head in frame the entire time. Feet cut off at the bottom is the most common problem with phone-filmed trends.
• One continuous shot. No jump cuts, no transitions. Cuts can make the output end early.
• Locked camera. A phone on a tripod or propped on a shelf. Camera moves can truncate the result too.
• Moderate tempo. Kling warns that very fast or complex motion can produce a shorter output. Slow the routine a little when you film it.
• Big moves, little travel. Kling recommends a wide range of motion with minimal movement across the floor. Arm waves and body rolls, yes. Running across a room, no.
• 3 to 30 seconds. Most trend dances fit in 8 to 15 seconds.
• Resolution within limits. Short edge at least 340 px, long edge at most 3850 px. Any modern phone clip is fine.

The best reference is one you film yourself. You control the framing, you own the recording, and you can slow the tempo slightly for cleaner results. Wear fitted clothing so the model can read your limbs, and film against a plain wall.

Photo requirements

• Full body, head to feet, if the dance is full body. A headshot cannot do footwork.
• One person, standing, facing the camera or close to it.
• Arms and hands visible, not in pockets, not behind the back, not holding a phone.
• Nothing blocking the body. No bags across the torso, no pets in arms, no table in front. Kling and fal both say the character should be unobstructed, and Oakgen's form says the character should take up more than 5% of the image.
• Space around the body. Kling says big motions need room in the image. Leave some wall above the head and on both sides.
• Same orientation as the clip. Vertical photo for a vertical dance clip.

If your only photo is a close-up, you can generate a full-body version of a character on the AI image generator first, then animate that. For real people, use a real full-body photo taken with their permission.

Method 2: image-to-video with a dance prompt

Any image-to-video model will try to make your photo dance if you ask. You write something like "the woman dances energetically to upbeat pop music, arms swinging, hips moving, camera static" and the model improvises.

It works for short, loose movement: a sway, a bounce, a spin. It falls apart for anything recognizable. There is no way to specify the eight counts of a trend in a prompt, and the model will not land beats of a song it cannot hear.

When to use it: quick, vibe-only clips, and subjects that motion control handles poorly. On Oakgen, Kling O1 is the budget option at 146 credits for 5 seconds, and Kling V3 Pro is 437 credits for 5 seconds (as of September 2026). The guide to animating a photo compares these and other image-to-video options in detail.

Prompt tips if you go this route:

• Keep it to one or two moves ("sways side to side, then raises both arms").
• Say "static camera" so the model spends its effort on the body.
• Ask for 5 seconds, not 10. Longer improvised dances drift.

Method 3: dance templates

Template tools apply a pre-built dance to your photo. You pick the dance, upload the photo, and wait. No reference clip, no prompt.

On Oakgen, Vidu Template in the AI video generator (image-to-video models) does this for a flat 78 credits per video as of September 2026. Its template list includes dance options such as subject_3, shake_it_dance, hip_twist, motor_dance, rat_dance, wave_dance, chill_dance, lit_bounce, and jiggle_up, plus a pet_belly_dance template and couple templates such as hug, couple_walk, and romantic_lift.

The trade-off is control. You get the dances on the list and nothing else, and you cannot change the timing. For a quick, fun clip, templates are often the smartest choice. For a specific trend that is not in the list, go back to Method 1.

Adding music without getting muted

Music is where most people get into trouble.

Safe ways to add music

1. Add the sound inside the platform. Upload your AI dance video without music (turn off Keep Original Sound in motion control), then add the song from TikTok's or Instagram's in-app library. The platform handles the license for eligible personal use.
2. Use music you own or generated. Make an original track on the AI music generator and edit it under the video before posting. Where you can publish that file depends on Oakgen's terms, the music and video providers' terms, and your rights in the inputs, including the music and the likeness of the person dancing. Read Oakgen's terms on commercial use of outputs before using it in paid ads, and check the royalty-free music guide for Reels and TikTok for the wider picture.

What the platforms actually say

• TikTok business accounts and ads: TikTok's help center says businesses cannot use the general music library for commercial use. They must use the Commercial Music Library, a pre-cleared library of over a million tracks, for organic posts, duets and stitches, ads, and branded content. TikTok's terms say those commercial sounds are licensed for use on TikTok only.
• Instagram and Facebook: Meta's Music Guidelines say using music for commercial or non-personal purposes is prohibited unless you have the licenses, and that unauthorized music can be blocked, muted, or removed. Meta also recommends shorter clips of music and always having a visual component.

The practical trap: people download a trending sound, bake it into the video file, then post the same file to Reels, Shorts, and a brand's ad account. That is exactly the use these policies exclude.

Timing the dance to the beat

Motion control copies timing from your reference clip, so the easiest way to land the beat is to film the reference while the song plays. Then line the generated video back up with the same track in your editor. For beat-matched cuts across several clips, the beat drop Reels workflow covers editing to the drop.

Ideas: pets, kids, couples, and characters

Pets. Kling's guide recommends real human motion and says a face very different from the reference, such as a cat face on a human dance, may degrade quality. So a human routine on a dog photo is a gamble. Better bets: the pet_belly_dance template, or image-to-video with a simple bounce or sway. For a stylized look first, turn the pet into a 3D character with a Pixar-style pet portrait, then animate the portrait.

Kids. Get a parent's consent (if it is not your own child), keep the dance age-appropriate, and think about whether it needs to be public at all. A private family chat is a fine destination.

Couples. Motion control animates one person per generation. Either generate each partner separately from the same reference and composite them side by side, or use a two-person template such as couple_walk or romantic_lift.

Characters and mascots. Motion control handles these well. A brand mascot, an illustrated character, or a figurine photo can perform a real routine. Kling says some stylized humanoid characters are recognized, so keep the character human-shaped with clear arms and legs. The Instagram Reels workflow shows how to turn clips like this into a posting schedule.

Grandparents and old photos. Popular and emotional, but ask the family first, since some love the result and others find it unsettling.

Consent rules for AI dance videos

• Only animate people who said yes, or yourself. Impersonating or deceiving with someone's likeness is banned in Oakgen's terms, and you must hold the rights and consent for every real person you upload.
• Never sexualized content of real people. Suggestive dance templates applied to someone else's photo cross this line fast. Oakgen's terms prohibit sexualized depictions of real people without explicit consent.
• Children need a parent's consent.
• Respect the reference creator. A trend clip belongs to the person who filmed it. The US Copyright Office says choreography can be protected, while social dances and commonplace moves may not qualify. For personal fun, credit the creator. For commercial work, film your own reference or get permission.
• Label synthetic people in ads. Do not present an animated person as a real customer.

Troubleshooting AI dance videos

Limitations

Oakgen's motion control animates one person per generation, has no resolution picker, and requires a paid plan. If you want a duet in a single pass, the templates are the only built-in route, and they limit you to the dances on their list.

Sources

• Kling AI, Motion Control User Guide, accessed 2026-09-26
• fal, Kling Video v3 Pro Motion Control, accessed 2026-09-26
• TikTok Ads Help Center, Commercial Music Library, accessed 2026-09-26
• TikTok, Commercial Music Library User Terms, accessed 2026-09-26
• Meta, Music Guidelines, accessed 2026-09-26
• US Copyright Office, Circular 52: Copyright Registration of Choreography and Pantomime, accessed 2026-09-26
• Oakgen Terms of Service, accessed 2026-09-26
• Oakgen model configuration for Kling V3 Pro Motion Control, Kling V3 Pro and Kling O1 image-to-video, and Vidu Template (limits, template list, and provider cost at 1 USD = 260 credits), checked 2026-09-26