RefVid
How-to

How to shoot product reference photos for AI video

Good product reference photos are sharp, evenly lit, shot against a plain backdrop, and cover the hero angle, the sides, the back and a close-up of the logo. A phone camera is enough. The goal is not a beautiful catalog image. It is a clear record of the product's shape, color and markings, so a video model can rebuild it in any scene. This guide covers the setup, the shot list and the changes for difficult materials.

Generated with RefVid on Kling 3.0 Omni from three plain-backdrop reference photos of the bottle, prompt: a slowly rotating pedestal in a dark studio.

The setup: light, backdrop, camera

1. Pick the light. Daylight from a large window, out of direct sun, is the easiest source. Place the product about a yard from the window and put a sheet of white card or foam board on the other side to fill in shadows. If you use lamps, use two of the same type and color temperature, one on each side at 45 degrees. Mixed light, such as a warm bulb plus daylight, shifts colors between angles, and the model learns the wrong color.

2. Pick the backdrop. A sheet of white or light gray paper curved up behind the product, so there is no horizon line, works for most items. Use mid-gray for white products so the edges do not disappear. Avoid patterned cloth, wood grain and anything with text. Video models treat backgrounds that appear in several references as part of the subject.

3. Set the camera. Use the phone's main lens, not the ultra-wide, which bends straight edges. Step back and zoom to 2x if your phone has it, so the proportions look natural. Turn off portrait mode, beauty filters and color effects. Tap to focus on the product and lock exposure. Shoot at full resolution and keep the original files.

The shot list

4. Shoot these images in order, without moving the lights. The hero angle: a front three-quarter view at the product's eye level, the way it would sit on a shelf. The straight front. One side. The back, including any label panel. A top-down view if the top carries a logo or a distinctive cap. A close-up of the logo or main marking, filling most of the frame and perfectly sharp.

5. Add one in-use image if the product is held or worn. A hand holding the bottle, the bag on a shoulder, the watch on a wrist. This shows scale, which product-only shots cannot, and stops the model from rendering a perfume bottle the size of a wine bottle.

6. Optionally record a short turntable clip. Rotate the product slowly on a lazy Susan for about eight seconds against the same backdrop. Some models accept a reference video. Kling 3.0 Omni accepts one of 3 to 10 seconds alongside up to four images, according to Kling's guide as of October 2026.

Difficult products

Shiny and metal surfaces reflect the room, and those reflections end up in the video. Move the white card closer to create one clean, broad highlight, and shoot from slightly above so the reflection is of the ceiling or card, not of you.

Clear glass and liquids lose their edges against white. Use a dark backdrop for one or two shots, or place a strip of black card just outside the frame on each side to draw dark lines along the edges.

Soft goods such as clothing and bags change shape. Stuff bags with paper so they hold their form, steam clothes, and keep the shape the same across all images.

Products with variants should be shot as separate products. A bottle in blue and the same bottle in red are two reference sets. Mixing colorways in one set produces a color in between.

Check before you upload

7. Review the set on a large screen, not on the phone. Zoom in on the logo. If it is soft at 100 percent, reshoot it; the model cannot recover detail that is not there. Check that colors match across the images. Crop out stray objects at the edges. Do not add text, watermarks or arrows.

Expect fine print to be weak whatever you do. Ingredient lists and small legal text are still redrawn by current video models. Plan scenes where the logo and shape carry the shot, and the small print is out of focus or turned away.

Common mistakes and checklist

The usual mistakes: shooting on a busy kitchen counter, using the ultra-wide lens for close-ups, catching your phone in a glossy surface, and mixing photos from different days.

Checklist: one light type; plain curved backdrop; main lens, no filters; hero, front, side, back, top and logo close-up; one in-use shot for scale; colors consistent; logo sharp at full size; one set per colorway.

To use the photos in RefVid, create a product and upload your best 1 to 5 of them: JPG, PNG or WebP, at least 300 px and up to 10 MB each. Put the hero angle first, because RefVid treats the first image as the main view. RefVid takes still images only, so keep the turntable clip for models that accept video.

Frequently asked questions

Can I use photos from my online store as references?
Often yes, if they are sharp, show several angles and sit on a plain background. Heavily retouched images with added props work less well. Add a plain logo close-up if your store photos lack one.
Do I need a professional camera?
No. A recent phone on its main lens, with soft window light and a plain backdrop, works well. Light and a sharp logo matter more than the camera.
Why does the label text come out wrong in the video?
Small printed text is the weakest point of current video models. Keep logos large and sharp in your references and avoid shots where fine print is the focus.
Which photos should I pick for a RefVid product?
Up to five: hero angle first, then side, back, logo close-up and an in-use shot. A 5-second test costs 210 credits, and the first 50 accounts get 630 free credits. Later accounts get 126.