AI Image to Video: Four Methods, and Which You Get
Four things people mean
| Method | What the picture does | You give | Here? |
|---|---|---|---|
| Animate a still | The picture is frame one; a sentence says how it moves | One image and a line about the motion | No. There is no one-button animate-this-photo tool |
| Pin first and last frame | Two pictures fix where a clip starts and where it ends | Two images | No. A segment is made from the cast pictures and its own storyboard text, not from frame pictures you supply |
| Pictures as references | The pictures say who and where; the model draws a new shot of them | One picture per character, location or prop | Yes. It is how every shot is made, and your own photo can replace any of them |
| Photos into an ad | Up to four photos are read, a product is redrawn, a spot is written around it | One to four photos and one line | Yes. 15, 30 or 60 seconds |
The difference decides what you see. With a reference, the picture says what a person or place looks like, not how the shot begins. The model composes a new picture of them from the shot text, close or wide, standing or sitting, so the result is your subject in a new shot, not your photo moving. If you need the exact frame to be your photo, that is the first method, and this tool is not it. Image to video AI covers the two routes that are offered in more detail.
What you have, and what to do with it
| You have | Do this | Expect |
|---|---|---|
| A product photo, or up to four of the same thing | Ad Video tab; the first photo is the subject | A made-up spot around your product, redrawn as a clean studio shot, ending on a card with your brand |
| A shop, restaurant, car or house | Ad Video tab | The place is kept as it is, people are removed, and the first scene is built from your own photo |
| A room or street a story should happen in | Upload & Replace on the location card; no people, 16:9 | Every shot at that location is built on it |
| An object that matters to the plot | The prop card, the object alone on a plain background | It is the reference whenever the prop is in frame |
| A character you drew or made | Upload & Replace, then Use this image as is | It becomes that character’s sheet untouched |
| A photo of a real person | The character card, with care | It may be flagged if it reads as a real person’s face; celebrities and brand mascots are refused |
| An app or website screenshot | Ad Video tab; one screen, no faces | Never redrawn, so the interface stays readable |
| One still you want to see move | Not offered | Closest: put the subject on a card and write a one-shot scene around it |
Putting your own picture on a card is a membership feature on the current plan settings; photos for an ad are a separate door and are not part of that. Use your own photo · Which photos work for an ad.
What happens to a picture when you pick it
Your browser reads WebP, HEIC, PNG or JPEG, flattens any transparency onto white and re-encodes it, so what arrives is always JPEG or PNG. The only refusals are files your own browser cannot open, such as SVG, camera RAW and HEIC outside Safari, and you find out the moment you pick the file. An ad takes up to four photos of the same subject, up to ten megabytes each and sixteen in all, and reads them once, for free, before anything is charged. The first time you upload anything, a window asks you to confirm that you have the right to use it and that anyone whose face or voice is in it consented. What you can feed in.
On a character card you choose how the picture is used: Optimize into a reference sheet redraws it in the project’s look on a clean background and is priced like an image, while Use this image as is is not an image generation. Either way, new videos for that character look at it from then on; clips made earlier are untouched. Your own reference images: what changes.
The same picture, three different shots
Because the shot text decides the framing and the action, one photo on a card gives you as many different shots as you write. With a character called Ana on a card:
| Shot text | What it asks for |
|---|---|
| [Shot 1] 3s: Close-up on @Ana at a window. She lets the curtain fall. | Her face large, one small action |
| [Shot 2] 5s: Wide shot, a hallway. @Ana walks away from the camera and stops at a door. | The whole figure in a room, moving away |
| [Shot 3] 4s: Over-the-shoulder, @Ana facing a stranger. @Ana: You are early. | A second person in frame and a spoken line |
A segment of shots like these is one generation, charged per attempt, and every take is kept. Editing the text is free, so write the shot until it says what you want, then generate. How to write one shot.
Across clips: what carries over from a picture
Each segment starts from the cast pictures and its own storyboard text. It does not start from the last frame of the segment before, so the same photo gives the same person in every segment, but a segment does not open on the exact picture the last one ended on. What has to carry across a cut, a bandage, a coat taken off, is kept by writing it as an event in the shot where it happens, or by a state picture for something that lasts. Why the next segment opens like the last one ended · What the system keeps track of.
When the picture you start from is a video
A finished clip can itself be the starting point. AI visuals makes the clip again from itself with one change from a fixed menu of eight, and everything else about how it was shot carries over. That is video to video, not image to video, and it is priced and limited differently. AI visuals: what it can change.
FAQ
Can I turn one photo into a video here?
As an ad, with up to four photos of one subject. A photo can also become a character, location or prop that shots are made around. There is no tool that animates a single still on its own.
Can I choose the first and last frame?
Not as pictures. You can describe how a segment opens in its first shot, but a segment is made from the cast pictures and its own storyboard text.
Will the video look exactly like my photo?
A product is redrawn as a clean studio shot, a place is kept, and a character sheet made from a photo keeps the face, hair and clothing redrawn in the project’s look.
What files can I upload?
Your browser reads WebP, HEIC, PNG or JPEG and sends JPEG or PNG. SVG and camera RAW are refused when you pick them.
How many photos can an ad use?
Up to four of the same subject, ten megabytes each and sixteen in total.
Does the same photo give the same person in every shot?
Yes. Every shot that shows the character is made from the same entry. Outfit changes are explicit states you choose.
Can I use a photo of myself?
A photo that reads as a real person’s face may be flagged, and uploading to a card is a membership feature. The guide on using your own photo explains the details.
What does it cost?
Reading ad photos is free. Video is 10 credits a second on the default tier, and the price shows on the button before you press it.
Start from the picture you actually have
A product, a place, a face you drew: each has a door.
Updated November 2, 2026