# Product Photo Checklist for AI Video: 12 Points

> What makes a product photo animate well in image to video tools, with a 12-point checklist, a phone shooting setup and a table of fixes for common problems.

URL: https://advideolab.com/blog/product-photo-checklist-for-ai-video
Published: 2026-10-13
Updated: 2026-11-03
Category: Guides
Author: The AdVideoLab team
Image: https://advideolab.com/blog/product-photo-checklist-for-ai-video.webp

When we trace a weak video back to its cause, the outcome was usually settled before anyone pressed Generate. The brief is fine. The photo is a 600 pixel crop of a supplier listing, with a watermark in one corner and the product's cap cut off by the frame edge. The model does what it can, which is to guess.

An image to video model treats your photo as the truth about the product. Everything it can see clearly, it tends to keep. Everything it can't see, it invents, and invented details are where labels turn into squiggles and bottles grow an extra ridge. So the fastest way to better AI video is usually a better source photo, and that takes about twenty minutes with a phone and a window.

This is the checklist we use ourselves, plus the fixes for the problems that show up most.

## Why the photo sets the ceiling

A video model has to produce every frame of the clip, each one consistent with your product. When the camera moves, it needs to know what the edges look like, how the light sits on the surface and what the label says. A sharp photo gives it that. A soft or cluttered photo forces it to fill gaps, and it fills them differently from frame to frame, which is why weak photos produce products that seem to shimmer or melt slightly as the shot goes on.

Our [image upload guide](https://advideolab.com/help/uploads-formats) puts it plainly: the model can only keep what it can see clearly. You can write a perfect brief and still lose to a blurry label.

## The 12-point checklist

Run through this before you upload. If you can tick ten or more, you're in good shape.

1. **Resolution.** At least 1500 pixels on the short side is our comfortable minimum for close-ups. AdVideoLab accepts anything from 300 to 6000 pixels per side and up to 25 MB, but 300 pixels will not survive a push-in.
2. **One product per frame.** If there are three products, the model has to decide which one is the hero. Shoot the bundle separately, or use Reference mode (more on that below).
3. **Clean edges.** The outline of the product should be clearly separated from the background. Dark bottle on a dark table is the classic failure.
4. **Label legible at phone size.** Zoom your photo to the size of a phone screen. If you can't read the brand name, the model won't hold it either.
5. **Whole product in frame.** No cropped caps, cut-off handles or missing corners. Leave a margin of roughly 10 percent around the product on all sides.
6. **The right angle.** A three-quarter view (turned about 30 to 45 degrees) shows front and side at once and gives the camera something to move around. Flat lays suit flat things: skincare sets, stationery, clothing laid out.
7. **Plain or simple background.** White, light grey, a wooden table, a bathroom shelf with little on it. Busy patterns bleed into the edges.
8. **Soft, even light.** Window light on an overcast day is ideal. Avoid on-camera flash.
9. **Controlled reflections.** Glass, chrome and glossy plastic should show a soft highlight, not a hard white blob or a reflection of you holding the phone.
10. **Colour true to the real thing.** No heavy filters or warm edits. If the jar is pale green in life, it should be pale green in the photo, because the video will copy it.
11. **No text overlays, stickers or watermarks.** The model tries to animate them. Supplier watermarks can also create rights problems.
12. **Natural shape.** Roughly square to portrait. Very long panoramic images or thin strips lose detail when they're fitted to a vertical 9:16 frame.

## Angle: three-quarter or flat lay?

We default to three-quarter for anything with depth: bottles, boxes, shoes, gadgets, bags. It tells the model what the side looks like, so when the camera orbits a little it has real information to work from.

A straight-on front shot is fine for a slow push-in, but it's risky for any movement that reveals the side, because the model has never seen the side. Our AI brief tells the model not to show sides that aren't in the photo, and that helps, but a three-quarter photo removes the problem instead of working around it.

Flat lays work well for products that are already flat and for sets where the arrangement is the point. They make good Single image videos with a slow overhead drift. They're a poor base for Creator videos, because a person can't naturally pick up an object the model has only seen from directly above.

## Glass, metal and other reflective products

Reflective products are where most home shoots go wrong. Three things help:

- **Move away from the window, not towards it.** Stand the product a metre or so back so the light is soft and broad instead of a bright rectangle on the glass.
- **Use a white card on the opposite side.** It fills the shadow and gives the glass a clean vertical highlight instead of a dark edge.
- **Check for yourself in the reflection.** Chrome caps and mirrored packaging will show you, the phone and the room. Shoot from slightly above, or hang a white sheet behind you.

For clear glass with liquid inside, a light background behind the bottle shows the colour of the liquid. A dark background hides it and the model may render an empty bottle.

## A phone setup that costs almost nothing

You don't need a studio. This is the setup in our cover photo and it's what we'd tell a friend to use:

1. **A table next to a large window**, ideally one without direct sun. Overcast days are your friend.
2. **White paper or a sheet of foam board as the background**, curved up the wall behind the product so there's no visible horizon line.
3. **A second white card** propped on the side away from the window to fill shadows.
4. **Phone on a small tripod or a stack of books**, lens at roughly the height of the product's middle for a three-quarter shot, slightly above for a flat lay.
5. **Tap to focus on the label, then lower exposure a little** so highlights on the packaging don't blow out.
6. **Use the main camera, not the ultra-wide.** Ultra-wide lenses stretch products near the frame edges. Step back and crop instead.
7. **Shoot five or six versions**, turning the product a few degrees each time. Pick the sharpest.

HEIC files straight from an iPhone are fine; we convert them. If one fails to read, export it as JPG and upload again.

## Fix it table

| Problem | What you'll see in the video | Fix |
| --- | --- | --- |
| Low resolution or soft focus | Label becomes a blur or fake lettering, edges shimmer | Reshoot at full resolution, tap to focus on the label |
| Product cropped at the edge | The model invents the missing part (a new cap, a different handle) | Leave space around the product on all sides |
| Several products in one shot | Products merge, swap places or multiply | One product per photo, or use Reference mode |
| Busy background | Background details bleed into the product outline | Plain paper or a simple surface behind the product |
| Hard flash or window glare | A white hotspot that moves oddly or sticks to the frame | Diffuse light, move the product back from the window |
| Reflection of you in chrome | A ghostly figure appears in the product surface | Shoot from above or hang a white sheet behind you |
| Heavy colour filter | The product arrives in the video the wrong colour | Shoot neutral, no filters |
| Watermark or text overlay | The text animates, warps or stays glued in place | Ask the supplier for clean images |
| Front-only photo, camera orbits | The side or back is guessed and looks wrong | Three-quarter angle, or keep the move to a push-in |
| Ultra-wide lens | Product looks stretched or bulging | Main lens, step back, crop later |

If a video still comes out wrong after all this, our [failed videos guide](https://advideolab.com/help/failed-videos) covers refusals and odd results, and credits are refunded automatically when a render fails.

## When one photo isn't enough

Some products can't be shown in a single frame: a three-piece skincare set, a jacket where the lining matters, a lamp that looks different on and off, a product that comes in four colours. That's what [Reference mode](https://advideolab.com/help/reference-mode) is for. You upload several photos and the model builds a scene that stays true to all of them.

On Seedance 2.0 you can attach up to 9 photos; on Seedance 2.5, up to 30, though 2.5 refuses photos with real human faces in this mode. A few habits make Reference sets work better:

- **Put the main product photo first.**
- **Keep the lighting consistent** across the set, so the model isn't averaging a warm photo with a cold one.
- **Say what each photo is for** in the brief, for example "photo 3 is the texture of the cream, use it for the close-up".
- **Add one setting photo** if you want a particular place, like your own bathroom shelf.

For a simple product with one good photo, stay with Single image or Creator mode. Reference adds control, and it also adds more for the model to reconcile.

## Match the photo to the mode

| Mode | Best source photo |
| --- | --- |
| Creator (a person holds and talks about it) | Three-quarter view, whole product visible, clear shape a hand can grip |
| Single image (product on its own) | Highest resolution you have, clean background, strong label |
| Transition (before and after) | Two photos with matching angle and framing |
| Reference (several photos) | A consistent set, main product first |

The [examples page](https://advideolab.com/examples) shows what finished videos look like across the four modes. Our [image to video ads page](https://advideolab.com/image-to-video-ads) covers the Single image flow in more detail, and our guide to [writing video prompts that sell](https://advideolab.com/blog/how-to-write-video-prompts-that-sell) picks up where this checklist stops.

## FAQ

### Can I use photos from a supplier listing?

Yes, if they're high resolution, have no watermark and you have the right to use them. Many suppliers will send clean originals if you ask. If every listing image is watermarked or tiny, ask the supplier before you spend credits on a video that will come out soft.

### Should I remove the background before uploading?

Not usually. A clean, real background often looks more natural than a cut-out on pure white, and cut-out tools can leave rough edges that the model then animates. If your background is busy, reshooting on plain paper tends to beat editing.

### Does a higher-resolution photo cost more credits?

No. Credits depend on the length and output resolution of the finished video, not on the photo you upload. A 15 second Seedance 2.0 video at 720p costs 330 credits with a 1500 pixel photo or a 5000 pixel one.

### My product is tiny, like a ring or earbuds. Any special tips?

Fill more of the frame with the product and use the main lens up close rather than zooming digitally. Small products lose detail fastest, so the 1500 pixel minimum matters most here. For Creator videos, a hand for scale in the brief helps the model get the size right.

### Is one great photo better than several average ones?

For Single image and Creator mode, yes. One sharp three-quarter photo gives better results than a mediocre photo with extra angles. Reference mode is where several photos pay off, and even there a consistent set beats a large, mixed one.

If you want to test your photo before committing to a plan, the one-time $3.99 [Trial Pass](https://advideolab.com/trial) gives you 170 credits at 480p, which is enough for one watermarked 15 second video.