Content / workflow
Text to video or image to video: which one should you use?
The difference is not quality. It is what you already have — a sentence or a frame — and how much control you need over the first second.
Every generation in VideoMaker AI ends in the same place: a short clip, quoted in credits before it runs, reserved while it renders, settled only when the file is delivered. What changes is where the shot starts.
Text to video starts with a description. Image to video starts with a photograph that becomes the first frame. Neither mode is "better". They answer different questions.
Text to video: the scene does not exist yet
Use text when you are still deciding what the shot should be. You can name the subject, the action, the camera, the light and the texture in one paragraph, then rerun a variation without reshooting anything.
It is the right mode for:
- a mood or visual idea you want to explore;
- a scene with no existing reference;
- a shot where framing and atmosphere matter more than a specific face or product.
The trade-off is that every detail is a suggestion. If the exact arrangement of a room matters, words will fight you.
Image to video: protect the first frame
Use an image when the opening frame is already right: a product photo, a storyboard panel, a still from a previous shoot, a character you have designed. The photo becomes the first frame, and the prompt describes how that frame should move.
The image locks composition, color and identity. Your prompt should describe motion and camera, not a new scene. "Slow push in, dust drifting through the window light" works. "A different room with a chair" does not — the model has to keep the frame you gave it.
A few practical rules:
- One image per job. JPG, PNG or WebP, up to 8 MB.
- The image is held in private storage and deleted when the job reaches a final state. It is never used as a public sample.
- Auto aspect ratio follows the source frame. Choose an explicit ratio when the delivery needs a different frame.
- Keep the first action small. A 4–10 second clip has room for one gesture, not a sequence.
The parameters are shared
Both modes use the same controls and the same quote engine:
- Duration — 4 to 15 seconds, in whole seconds. Default 10.
- Resolution — 480p, 720p or 1080p.
- Aspect ratio — Auto, 16:9, 4:3, 1:1, 3:4, 9:16 or 21:9.
- Prompt — up to 2,000 characters. Image-to-video prompts are optional.
Credits are charged per second and scale with resolution: 5 per second at 480p, 11 at 720p, 27 at 1080p.
| Shot | Credits |
|---|---|
| 4s at 480p | 20 |
| 10s at 720p | 110 |
| 15s at 1080p | 405 |
Those are quote values, not estimates. The quote is pinned to the prompt, the image, the options and the pricing version. Change any of them and it becomes a new quote.
How to choose in one sentence
If you can describe the shot but cannot draw it, start with text. If you already have the frame and need it to move, start with an image. If the motion itself is the product — a fixed effect you want to reuse — start from an effect template instead.
Try both before you commit
Open the studio and describe a scene or upload a frame. The quote appears before you confirm; credits are reserved only when you press generate, and settled only when the video is delivered. A failed or refused job releases the full hold.
Most people learn the difference in one afternoon: text is for finding the shot, image is for keeping it.