What happens to a product photo before an ad is made
Three tools in this register publish what an input image may weigh or what using one does to the output. One generator publishes a step before any of that: the photo is read for category, product name, brand name off the signage and selling points, and people in a shop or car photo are removed. As of 2026-10-13.
| Item | What is published | Detail |
|---|---|---|
| HeyGen | An image up to 50MB | No reading step on the page read |
| Runway | 16MB for an image by URL, 5MB as a data URI | No reading step on the page read |
| SceneMixer | One to four JPG or PNG photos under 10 MB; the first is the subject | Read for category, name, brand and selling points; people removed |
| Veo | Reference images accepted; the clip is then 8 seconds | No reading step on the page read |
Inclusion rule. Tools whose own page states a limit or a consequence for an input image, or a step that reads the image before generating. A tool that only lists image-to-video as a mode does not earn a row. Order. Alphabetical by tool.
1A file ceiling is a limit; a reading is a brief
An upload ceiling tells an advertiser whether the photo will be accepted. It says nothing about what the model will make of it, which is the question that decides whether the ad is about the right product.
A reading step answers that before any money is spent on video: it names the category, the product and the brand it found, and lets the name be corrected before anything is made. One tool here documents that step and states that it costs nothing.
2People in the photo are a rights question
A storefront or a car photographed in the street usually has passers-by in it. One generator removes them as part of the reading, and refuses a screenshot that shows a real face, which keeps an identifiable stranger out of an advertisement nobody asked them to appear in.
None of the file-limit pages raises the subject, which is consistent with their purpose: they describe what the API accepts, not what an advertiser should send it.
3The first image decides the frame
Where several images are accepted, order matters. One guide states that the first image is the thing being advertised and fills the frame, while later images show angles, details or the place it lives. Using a reference image elsewhere fixes the clip length, which is a different kind of consequence: the input changes the output's shape rather than its subject.
- Upload ceilings100MB for a video, 50MB for an image or an audio file, 32MB for an asset upload
- Ceilings on what goes in16MB for an image and 32MB for a video by URL, 5MB and 16MB as a data URI, 200MB for an ephemeral uploadnot on what comes out
- Veo 3.1 and Veo 3.1 Fast8 seconds, 6 seconds, 4 seconds, with 8 seconds only if 1080p or 4k or using reference images
- What goes inOne to four JPG or PNG photos under 10 MB each; the first image is the thing advertised and fills the frame
- What happens to the photo firstThe photo is read before anything is made, returning category, product name, brand name off the signage and selling points, at no charge
- Photos with people in themPeople in a shop or car photo are removed, and a screenshot showing a real face is refused
4Sources
Format requirements come from Google Ads Help, read 2026-09-12, and tool limits from each vendor page listed on published limits, read 2026-09-22 and 2026-10-13. The rule against mixing the two is on method. Nearby: TikTok specs, Shape by platform, File limits.