
- Credits per clip
- 300
- Total credits
- 600
- Clips covered by one 600-credit pack
- 2
- 600-credit packs needed
- 1
A failed generation does not consume credits, but a retry is another full-cost job: budget the retries you expect, not the clips you hope for.
There are three ways into a dance clip on this studio, and they are not interchangeable. Each one takes a different thing from you as input, and each one fails differently. Picking the wrong one is the most expensive mistake available, because it costs a generation to discover it.
The three inputs
Text. You describe the subject and the movement. The model invents the person. Cheapest, no photo needed, no consent question — and the face will not be anyone you know.
A still photo. You supply the subject; you describe the movement. The face is preserved, which is the entire point, and the input quality decides whether it survives.
A reference video. You supply the movement, plus a still image for the subject. The model transfers one onto the other. The most control over the dance itself, the least tolerance for a bad reference.
Cost and risk, side by side
| Text to video | Image to video | Motion control | |
|---|---|---|---|
| Input | Words | Photo + words | Reference clip + image |
| Credits (5s / per mode) | 300 | 300 | 500 standard, 1,000 pro |
| Face is a specific child | No | Yes | Yes |
| Movement control | Prompt only | Prompt only | Direct, from the reference |
| Needs a photo you can use | No | Yes | Yes |
| Main failure mode | Generic subject | Identity drift | Bad reference → bad output |
| Consent question | None | Yes, if the child is real | Yes, plus the reference's rights |
Read the last three columns together: motion control has the most control and the most obligations. It needs a photo you can use and a reference clip whose movement you have the right to copy. That is two separate rights questions before you press generate.
Decision tree
Does the clip need to be a specific child?
- No. Use text to video. 300 credits, no photo, no consent question, no face to drift. This is the right answer more often than people expect — the format works on a generated subject.
- Yes. Go to the next question.
Do you already know the exact movement you want?
- No, "something dancey" is fine. Use image to video with a mechanical prompt: "two side steps, then a shoulder bounce, arms bent and low". 300 credits.
- Yes, a specific move. Use motion control with a reference clip. 500 credits for standard, 1,000 for pro.
Does the clip need to be sharp on a large screen or in a paid placement?
- No. Standard motion control, 720p, 500 credits.
- Yes. Pro, 1080p, 1,000 credits — and note that a single pack of 600 credits does not cover it.
Why text mode is underrated
The instinct is that a "real" photo makes a better video. Often the opposite is true: a generated subject has no identity to lose. Identity drift — the failure that ruins most of these clips — is a property of trying to preserve a specific face across frames. Remove the specific face and the failure mode disappears.
Text mode is also the only mode with no consent question, which makes it the only mode you can use freely for a commercial or public post involving a child.
What text mode gives up is the surprise of recognition: "that is my nephew" is a real part of the appeal for a family clip. If that is the point, pay for it in image mode and accept the drift risk.
Why motion control is overbought
Motion control is the mode people reach for when the result is disappointing, because "more control" sounds like "better". It is not a quality upgrade; it is a different input contract. Two things follow:
- A bad reference produces a bad clip, faithfully. The model is doing what you asked. If the reference has a shaky camera or a cropped subject, the output inherits it.
- It costs more than the entry pack at pro. 1,000 credits against a 600-credit pack means two packs for one clip. That is a decision to make deliberately, not as a retry.
The right time for motion control is when you have a specific move in mind and a reference you are allowed to use — for example, your own recording of the dance you want the child to do. That is also the most defensible version: you own both inputs.
Mixing modes in one project
The cheapest way to make a ten-second clip with two distinct movements is not a ten-second generation at 600 credits. It is two five-second generations at 300 each, cut together — same total cost, less drift per segment, and you can discard one if it fails.
The same logic applies across modes: a text-mode opener plus an image-mode close is a normal edit. Nothing requires a project to use one mode throughout.
A five-minute plan before spending
- Decide whether the child must be specific. That single answer eliminates one mode.
- If specific, check the photo: sharp, front-facing, whole head, no occluding shadow. If it fails, fix the photo first — a rejected job costs time, a producing job with a bad photo costs 300 credits.
- Write the movement mechanically. If you cannot, and you have a reference you can use, choose motion control instead of guessing at words.
- Pick the resolution by where the clip will be watched, not by what sounds better.
- Price it:
The planner above is the same table the studio deducts from. Set the mode, set the duration, and read the number before you commit.
The one-line version
Text mode when the child need not be specific, image mode when it must, motion control when you already own the movement. Everything else — duration, audio, resolution — is a cost decision layered on top of that choice, and the planning guide does the arithmetic.
If the mode is right and the clip still comes back wrong, the defect points at the input rather than the choice: the failure guide maps each symptom to the fix. And if the child is real, the consent guide comes before everything above.
Frequently asked questions
Which mode is cheapest?
Text to video and image to video at five seconds without audio, at 300 credits each. Motion control costs more: 500 credits for standard 720p and 1,000 for pro 1080p, because it runs a different model that transfers movement from a reference clip.
When should I use a photo instead of text?
When the clip needs to be a specific child. Text to video generates a subject that resembles nobody in particular; image to video starts from your photo and preserves that face, which is what you want for a birthday or family clip — and it is also why consent matters.
What does motion control actually do?
It takes the movement from a reference video and applies it to a still image. You are not describing the dance in words; you are supplying it. That makes it the right mode when a specific move matters, and the wrong mode when you do not have a reference clip you are allowed to use.
What is the difference between standard and pro motion control?
The render resolution and the credit cost. Standard returns 720p for 500 credits; pro returns 1080p for 1,000. The movement behaviour is the same, so choose pro only when the extra sharpness is visible where the clip will be watched.
Can I combine modes in one project?
Yes, and it is usually the best plan: generate silent clips in whichever mode suits each shot, then cut them together in the editor. Mixing modes costs nothing extra and avoids paying for a long generation just to get a second movement.
![50+ Best AI Baby Dance Video Prompts That Actually Work [2026 Guide]](/imgs/blog/baby-dance-ai-prompts.webp)
![AI Baby Dance Video Generator: How to Make Your Baby Dance with AI [2026]](/imgs/blog/how-to-make-baby-dance-ai-video.webp)







