How to choose a photo for image to video.
Image-to-video models keep what they can see and invent what they can’t. The right photo is the difference between a believable clip and wasted credits.
Why the photo matters
The model uses your photo as the first frame and the reference for everything that follows. Detail that isn’t in the photo — a hidden label, a face in shadow — has to be invented, and invented detail is where AI video starts to look wrong.
The checklist
- 01Sharp and in focus
Tap to focus on the face or product before you take the photo.
- 02Even daylight
Face a window. Avoid a strong flash and deep shadows.
- 03Face fully visible
Both eyes visible, nothing covering the face.
- 04Label facing the camera
Hold the product so its name is straight and readable.
- 05Space around the subject
Leave room for the hands and product to move.
- 06The right shape
Vertical photo for 9:16 video, wide photo for 16:9.
Describe the motion simply
One or two actions per clip: “she raises the bottle to the camera and smiles”, “he turns the phone to show the back”. Several actions in a short clip make movement rushed and unnatural — the Director warns you when a brief has too many.
Common problems and fixes
| Problem | Likely cause | Fix |
|---|---|---|
| Face changes during the clip | Face small, in shadow or turned away | Use a closer, front-facing photo in daylight |
| Label text is blurry or wrong | Label not readable in the photo | Retake with the label straight to the camera |
| Hands look strange | Too many hand actions | Ask for one simple action |
| Movement looks stiff | Subject cropped tightly | Use a photo with space around the person |
| Looks artificial | Beauty filter or heavy edit on the photo | Use the unedited photo |
Permission
Only animate photos you have the right to use, with consent from every person shown. See the Acceptable Use Policy.