AI cover art
Generating a cover image from a prompt — what it costs, what you get back, and how to work with the result.
The Cover Studio can generate an image from a text prompt and drop it in as a front-panel image layer.
2 credits per image, on paid plans.
Using it
Describe the image you want. You get back a picture on a layer, which you then move, scale, crop and set type over like any other image.
The model is fixed, and shown as a read-only field in the studio — there is no model picker here, because there is one image model.
Working with the result
The result is artwork, not a cover. The layout is still yours:
- Put the title and author on their own text layers, above the image. Do not ask the model for lettering — image models render text unreliably, and type on its own layer stays crisp at print resolution and re-themes with your design.
- Check it against the barcode zone if it extends onto the back panel.
- Watch the resolution. The release board checks image DPI against your trim, and a generated image stretched across a large wrap can fall below the print floor. Use it on the front panel at its natural size, or as one element of a composition, rather than blowing it up across the whole cover.
Regenerating
Each generation is a new image and a new 2 credits. There is no variant history here, so if you like something, keep it — delete the layer only when you are sure.
AI disclosure
An AI-generated cover image is evidence for the AI content declaration on the release board, which most platforms now require.
The evidence is derived from the layers in your book, so deleting the AI image deletes the evidence. Declaring more AI use than can be demonstrated is always fine; declaring less is a blocker.
What to watch for
- A failed generation is refunded. Credits are reserved before the provider is called and returned in full if the call fails.
- Image generation is a paid-plan feature, like page art and research.
- Prompts steer, they do not guarantee. Negative instructions ("no text", "no border") are advisory to an image model and are sometimes ignored. Look at what came back before building a design around it.