AI isometric illustration generator: a matching set through an API
Make a set of isometric illustrations that match: one style anchor, fixed camera words, and a per-item prompt. Model, ratio and cost per image on Sume.

To get a set of isometric illustrations that look like one family, generate one anchor image you like, then pass it as a reference on every later call and keep the camera and style words identical. On Sume that is one POST /v1/images per illustration. Say "isometric view, 30 degree camera, flat colours, no perspective distortion" every time, since an image model has no isometric switch.
Fix the style first
Write a style block of about three lines: camera, lighting, palette, outline weight. Test it on three subjects before you commit. Then keep the winning frame as the anchor reference. The same trick is used for icon sets (icon sets from one anchor).
- Camera: "isometric, 30 degree camera angle, orthographic".
- Surface: "flat vector look, soft shadow below the object".
- Background: plain, one colour, same across the set.
- Ratio:
1:1for icons and cards,16:9for scenes.
Which model
The catalog has several that edit from references: GPT Image 2.5, the Nano Banana models, FLUX.2 pro, Seedream 5.0 Lite and others. Recraft V4 is text-to-image only, so it cannot take the anchor (Image API). Pick on how well your anchor style holds across five test subjects, not on a vendor claim.
| Model id | Takes references | List rate x 1.25, per image |
|---|---|---|
| openai/gpt-image-2.5 | Yes, up to 16 | Token based, check the endpoints call |
| google/nano-banana-2 | Yes | $0.10 at 1K |
| black-forest-labs/flux.2-pro | Yes | $0.04 |
| bytedance-seed/seedream-5-lite | Yes | $0.05 |
| recraft/recraft-v4 | No | $0.05 |
Cost of a set
Image calls bill per completed image at list rate times 1.25, rounded up to a cent. Twelve illustrations on FLUX.2 pro at the table rate is about $0.48 before retries. Failed generations are not billed. The live number for your account is in GET /v1/images/models/{id}/endpoints under pricing.
Sources
Related posts
More in Use cases
- AI meme image generator API: make the picture, then set the caption
For a meme, generate the image without words and add the caption in code, or quote it for GPT Image 2.5 and check the spelling. A Sume request.
- Ad music cutdowns: 15, 30 and 60 seconds from one AI track
Make 15, 30 and 60 second versions of one AI music bed: three prompts or one cut with a fade-out. Costs and limits from Sume's Music Router and Timeline docs.
- AI music for a tribute slideshow: a gentle bed for a memorial video
Score a tribute or memorial slideshow with AI music: an instrumental prompt, stills as timed holds, fades, and a listen-through before the video is shared.
- AI newsletter header image: generate a 4:1 banner and crop to fit
Newsletter headers are wide strips. Ask Nano Banana 2 for 4:1 or 8:1 on Sume, keep the subject in the middle, and crop to your template. Request and crop math.
Written by Sume