블로그로 돌아가기

Make a Photo Dance with AI: Beginner Tutorial (Do's and Don'ts)

A step-by-step beginner tutorial for making a photo dance with AI: photo and reference video checklists, upload requirements, do's and don'ts, and quick fixes for common failures.

Sep 8, 2026Animaker DevAnimaker Dev
Make a Photo Dance with AI: Beginner Tutorial (Do's and Don'ts)

Making a photo dance with AI takes two inputs: a clear photo and a reference video that shows the movement. The AI maps the motion from the video onto the person in the photo. When the result looks bad, it is almost always one of the two inputs — not the tool.

This tutorial shows you exactly what to use, what to avoid, and how to fix the result when something looks off. Follow the two checklists and your first generation will already be close to what you pictured.

What You Need Before You Start

InputFormatSize limitLength
PhotoJPG, PNG, or WebPUp to 20 MB
Reference videoMP4, MOV, or WebMUp to 500 MBUp to 60 seconds

Keep both files on your computer before you start. A 10 to 30 second reference video is the sweet spot for a first test.

Step 1: Choose the Right Photo

The photo defines the person in your video. The clearer the photo, the more stable the result.

DoAvoid
Use a front-facing portrait with the whole face visibleGroup photos — the AI can only animate one person
Use even, natural lighting on the faceDark photos, strong shadows, or heavy filters
Keep the face unobstructed (no sunglasses, masks, or hair covering it)Faces at steep angles or turned away
Show the hands if your dance uses hand movementsCropping the hands out when the motion needs them
Use a full-body photo for leg or full-body movementsExpecting leg motion from a headshot
Use a plain, simple backgroundBusy backgrounds with objects in front of the person

One rule to remember: if you cannot see a body part clearly in the photo, the AI cannot move it well.

Step 2: Choose the Right Reference Video

The reference video is the movement instruction. It does not need to be cinematic — it needs to be readable.

DoAvoid
Use a video with exactly one personVideos with two or more people, even in the background
Keep the camera still (tripod, or a phone resting on something)Handheld footage that shakes or drifts
Show the complete action from start to finishActions cropped by the edge of the frame
Use even lighting and a simple backgroundBacklight, stage strobes, or heavy motion blur
Keep it 10–30 seconds for your first testA full 60-second choreography before you know the workflow

Good sources: your own phone recording of a simple move, a friend demonstrating a gesture, or a clean clip you have the rights to use.

Step 3: Upload and Generate

  1. Open the create page and upload your photo.
  2. Upload the reference video.
  3. Start the generation and wait — most tasks finish in about 10 to 15 minutes.
  4. Preview the result and download the MP4.

If the upload itself fails, it is almost always a spec issue:

ErrorFix
Image rejected or "too large"Compress the photo under 20 MB, or export as JPG
Video rejected or "too large"Trim the clip and export under 500 MB
Video rejected for lengthCut the reference to 60 seconds or less
Wrong formatConvert to MP4, MOV, or WebM

If a generation task fails, the consumed credit is refunded automatically.

Step 4: Check the Result — and Fix the Right Input

When the output is not what you wanted, change the input that caused it. Changing the wrong one wastes a generation.

What you seeThe problem is inFix
Face looks distorted or unlike the personPhotoUse a sharper, front-facing photo with better lighting
Arms or legs blurry or fusedReference videoPick a clip without overlapping limbs or occlusion
The action does not matchReference videoMake sure the full action is visible and uncut
Background flickers or warpsPhotoUse a plainer background with the subject clearly separated
The whole video looks unstableBothRestart with the simplest possible inputs: one static portrait, one 15-second clip

A good habit: keep every photo and video that produces a good result. They become your reusable starting point for the next video.

FAQ

What kind of photo works best for making a photo dance?

A clear, front-facing portrait with even lighting and an unobstructed face — JPG, PNG, or WebP up to 20 MB. If the dance uses hand movements, include the hands.

How long should the reference video be?

Up to 60 seconds, but 10 to 30 seconds is ideal for a first test. One subject, stable camera, complete action.

Why does my AI dance video look distorted?

Usually the inputs: a blurry or blocked face, a reference with overlapping limbs or motion blur, or a shaky camera. Fix the input before regenerating.

Can I use a group photo or a video with two people?

Not recommended. The AI animates one subject. Generate one video per person instead.

Do I need a full-body photo?

Only when the action involves legs or full-body movement. For gestures and upper-body motion, a chest-up portrait works.

Why was my file rejected?

Most likely the image is over 20 MB, the video is over 500 MB, or the video exceeds 60 seconds. Compress, trim, and use a supported format.

Why was my file rejected when uploading?

The three most common reasons are: the image is over 20 MB, the video is over 500 MB, or the video is longer than 60 seconds. Compress the image, trim the video, and use JPG, PNG, WebP, MP4, MOV, or WebM.

Try It

Pick one portrait and one 15-second reference clip, then make your photo dance with Animaker. For different subjects, try the dedicated flows for baby photo dance, pet photo dance, and old photo dance. To go deeper on motion inputs, read our reference action video tips.

관련 글