Making an image dance is an image-to-video task: you provide a still image, describe the movement, and let a video model create new frames around that subject. If you need a particular choreography, a reference-motion workflow adds a second input—a short video whose movement guides the result.
This guide explains both paths in the current Baby Dance AI interface. It does not promise a specific render time, perfect identity preservation, automatic music synchronization, or a watermark-free result. Those details can vary by mode, provider, queue, and output.
Choose the right workflow first
| Goal | Workflow | Inputs | Best when |
|---|---|---|---|
| Add a plausible dance from a written direction | Image to Video | One image and a prompt | The exact choreography is flexible |
| Follow movement from an existing clip | Motion Control | One image, one reference video, and an action prompt | Timing and body movement matter |
| Explore a scene without a source image | Text to Video | A prompt | You do not need to preserve a specific character |
Start with Image to Video if you are learning. Use Motion Control only when you own or are authorized to use both the source image and the reference clip.
1. Check rights, consent, and suitability
Before uploading anything, confirm that you have the right to use it. For a real child or any identifiable person, obtain appropriate permission from the person or their guardian. Do not upload sensitive personal information, private footage, or media intended to deceive viewers.
A usable source image generally has:
- a clearly visible subject;
- enough of the body in frame for the requested movement;
- stable lighting and reasonable sharpness;
- limited obstruction around the face, hands, and feet;
- a simple background when subject separation matters.
These are practical checks, not guarantees. A high-resolution photo can still produce artifacts, while a modest image can sometimes work well.
The current uploader accepts common image formats such as JPG, PNG, and WebP, with the interface showing a 10 MB limit. Follow the live form if those limits change.
2. Create a first version with Image to Video
- Open the AI Dance Video Generator.
- Select Image to Video.
- Upload an authorized image or provide a direct public image URL.
- Write a prompt that describes movement, camera behavior, setting, and pace.
- Open Advanced Settings to review the available duration, aspect ratio, and audio option.
- Check the displayed credit cost before submitting.
- Sign in if required, submit the job, and follow its status in the task interface.
Generation runs as an asynchronous job. You may leave the page after submission and check the activity view later. Queue time and output availability depend on the active provider.
3. Write a prompt the model can interpret
A useful dance prompt has four parts:
subject continuity + movement + camera + scene constraints
For example:
Keep the same illustrated character and outfit. Perform two gentle side steps, a small arm wave, and a short turn. Use a stable full-body camera, soft indoor light, and smooth motion. Avoid sudden cuts and extra people.
For a real family photo, keep the request age-appropriate and avoid wording that sexualizes, humiliates, or misrepresents the person.
Prompt details that usually help
- Movement: side step, hand wave, small turn, gentle bounce.
- Pace: slow, relaxed, energetic, or evenly timed.
- Framing: full body, medium shot, static camera, or slow push-in.
- Continuity: keep the same outfit, face, background, and lighting.
- Constraints: no scene cut, no extra people, no rapid camera shake.
Avoid a long list of conflicting directions. If the result is unstable, shorten the prompt and test one change at a time.
4. Use Motion Control for a specific dance
Motion Control adds a reference video. The generated character is guided by movement from that clip, but it is still a generative result rather than a frame-perfect copy.
- Select Motion Control or Motion Control Pro.
- Upload the character image.
- Choose an available dance template or upload an authorized reference video.
- Add a short action prompt that clarifies the intended subject and scene.
- Compare the available orientation, sound, duration, and cost controls.
- Submit only after the previewed inputs and displayed credit charge are correct.
Standard and Pro modes can use different providers, supported durations, costs, and controls. Treat the labels in the live form as the source of truth.
A better reference clip
Choose a clip where:
- one main dancer remains visible;
- limbs are not constantly hidden or cropped;
- the camera does not cut between angles;
- the motion fits the source image's visible pose;
- you have permission to reuse the choreography and footage.
Keeping original sound is a separate setting when supported. Do not assume a motion reference automatically transfers music or grants music rights.
5. Select duration, ratio, and audio deliberately
Short tests cost less than repeatedly submitting the longest option. Begin with the shortest available duration that can demonstrate the movement.
Choose aspect ratio for the destination:
- 9:16 for most vertical short-video layouts;
- 16:9 for landscape players and presentations;
- 1:1 where a square asset is required.
An audio switch means the active provider may generate audio; it does not guarantee a particular voice, song, beat match, or licensing status. Review generated audio before publishing and replace it in an editor when necessary.
The form calculates credits from the selected mode and options. Check that live number before every submission because provider pricing can change.
6. Review the result before publishing
Inspect the complete clip, not only its thumbnail. Check:
- face and identity consistency;
- hands, feet, clothing, and background edges;
- abrupt camera or scene changes;
- inappropriate or unsafe motion;
- audio artifacts and lip movement;
- rights, consent, and AI-disclosure requirements;
- the destination platform's current rules.
If a clip contains an identifiable child, consider whether publishing it creates an unnecessary long-term digital footprint even when you have permission.
Troubleshooting common failures
The body bends or limbs duplicate
Use a source image with more of the body visible. Reduce movement complexity, remove rapid turns, and test a shorter clip.
The face changes
Use a clearer front or three-quarter view, ask for a stable camera, and reduce extreme head rotation. No prompt can guarantee perfect identity preservation.
The character barely moves
Describe two or three concrete actions instead of using only “dance.” For precise choreography, try an authorized reference clip in Motion Control.
The upload fails
Confirm the file type and size shown by the uploader. If using a URL, use a direct public media URL rather than a private cloud-drive or social-sharing page.
The job stays pending
Provider queues can be variable. Use the task status view rather than submitting the same request repeatedly, which can create duplicate credit charges.
The result has a watermark
Watermark behavior depends on the provider, model, plan, and output. Review the watermark guide instead of assuming a paid plan changes every provider's output.
Privacy and safe handling
Browser conversion tools on the site process supported files locally, but hosted AI generation sends inputs to configured storage and generation providers. Do not confuse those two processing models.
Use only the minimum media needed for the job, remove unrelated personal details, and avoid uploading confidential footage. See the Privacy Policy for account, upload, provider, and retention information.
Frequently asked questions
Can one photo become a full dance video?
It can become a generated short clip, but the model invents frames that were not present in the photo. More complex or longer movement increases the chance of visual drift.
Do I need a reference video?
No. Image to Video works from an image and prompt. A reference video is useful when the movement itself is important.
Is it free?
No. New accounts start with 0 credits, so the first clip is paid, and every hosted job consumes the credit amount shown in the form. Only the in-browser free tools and the reading guides cost nothing.
Can I use any song or dance clip?
No. You are responsible for the rights to audio, choreography, footage, and identifiable people. A technically accepted upload is not proof of legal permission.
Which page should I use now?
Use the AI Dance Video Generator for image-to-video and reference-motion workflows. For a broader explanation of inputs and provider differences, read About Baby Dance AI.