AI Transparency
Our portraits are created with artificial intelligence and then printed. They are not hand-painted, and we don't want anyone to think they are. Here's what the AI does, what people do, and how to recognize AI-generated images.
Updated
In short. Every portrait is generated by a Google AI model, on Vertex AI, from the photo you upload. The sample portraits you see on the site are AI-generated too; the product photos, like the canvas on the wall or the poster, are real photos. We add the name on the nameplate ourselves, not the AI. And you choose which portrait gets printed.
What the AI does
When you upload a photo of your pet and choose a style (regal, watercolor, minimal line, or pop), we send the photo and the style instructions to a Google image generation model on Google Cloud's Vertex AI service. The model creates a new portrait of your pet in that style. We show you four previews, and you can ask for new ones up to two times.
The portrait is a generated image, then printed by a print lab on canvas or paper. It is not a hand-painted portrait.
What the AI doesn't do
- The name on the nameplate isn't written by the AI: our software adds it in a typeface, so the name is spelled exactly as you entered it.
- The final choice isn't made by the AI: we print only the preview you approve, exactly as it is.
- Memorial portraits: if the portrait honors a pet who has passed away, a person at Redbit also reviews the previews before we send them to you.
Likeness and limits
The AI tries to keep your pet's breed, colors, markings, and features, but it can get details wrong or interpret them. That's why you see the previews before you pay, and you only pay if you approve one. If none of them works for you, the order is canceled and you aren't charged.
Your photo
Your photo is used only to create your portrait. Google processes it on our behalf and commits not to use Vertex AI customer data to train its models. How we handle your photo and when we delete it is explained in our Privacy Policy.
Images on this site
| Image | How it's made |
|---|---|
| Sample portraits in the four styles | Generated with AI |
| Your previews and your portrait | Generated with AI from your photo |
| Product photos: the canvas on the wall, the poster | Real photos of printed products |
Provenance metadata
Images created by Google's model come with provenance metadata under the C2PA standard (Content Credentials) and an invisible Google digital watermark (SynthID) embedded in the image itself. When we add the nameplate or prepare the previews the file changes, and the original C2PA metadata does not survive. That's why, in the files we keep and in the digital file we deliver, we add our own IPTC metadata stating that the image is AI-generated. A print on paper or canvas, of course, can't carry metadata.
Why we tell you
Article 50 of the EU Artificial Intelligence Act (Regulation (EU) 2024/1689) requires AI-generated content to be identifiable. We'd do it anyway: a portrait of your pet is worth what it is, and we'd rather you know exactly how it's made.
Questions? Email info@coatandcanvas.com.