Projects & product types · Updated September 5, 2026
Choosing a generation model
Each product type lets you pick the AI model for your images. Cheaper models are faster and cost less; premium models give better composition and detail.
The model picker sits in the right sidebar of the create form. It’s a radio list, pick one. Each model shows its per-image credit cost inline so you can see the trade-off before you launch.
Where the picker appears (and where it doesn’t)
| Product type | Model picker |
|---|---|
| Wall Art | Yes, multiple models |
| Poster | Yes, multiple models |
| Clipart | Yes, multiple models |
| Greeting Card | Yes, multiple models |
| SVG | Yes, two quality levels: Standard and Premium |
| Custom | No, no image generation |
For SVG, the picker offers Standard (the default, best for flat shapes and bold designs) and Premium (drawn as vector from the start, best for lettering and intricate designs). For custom, the form omits the picker: you upload your own images, so there is nothing to generate.
What changes when you pick a different model
Only one thing changes: the model that generates your images. Everything else is identical:
- The same design-option fields are available (art medium, color palette, mood, room/setting, composition, negative prompts). There’s no longer a tiered system gating these, every model gets all options.
- The same brief and prompt LLM logic runs.
- The same upscale, processing, and delivery steps run afterward.
The old tier system (Creator / Studio / Signature) was retired in May 2026. The launch payload still includes a qualityTier field for backward compatibility but it no longer drives the UI.
Picking by intent
Without naming specific models (they rotate as the market changes, the picker always shows what’s currently available), the trade-off is:
- Cheaper models (low credit cost) are best for prototyping, batch volume, and styles where composition matters less than throughput, e.g. abstract patterns, simple clipart, mood-board pieces.
- Mid-tier models are the workhorse for most listings. Good composition, reasonable cost, predictable output.
- Premium models (high credit cost) are worth it for hero pieces where you need cleaner detail, better text rendering, or more reliable adherence to specific style cues (vintage botanical, fine-line illustration, specific photography styles).
If you’re unsure, start with a cheaper model on the first 1-2 pieces in a new direction. If the composition holds together, scale up; if not, you’ve spent less to learn it.
Credit cost is per set, not per image
Clipart is priced per image (5 credits an image on the default model, so a 20-image set is 100 credits). SVG sets are priced by design count (a 5-design Standard set is 35 credits). Other models change the price. The sidebar shows the exact total for the launch before you run; that is the number to look at.
Upscale and infrastructure models stay hidden
Only the generation model is user-facing. The upscaler, background-removal pass, and other infrastructure models stay out of the UI, you can’t pick or swap them. That’s intentional: the model behind a step like “background removal” is an implementation choice, not a creative one.
Related
- How credits are used per tool: the credit cost breakdown across a full run.
- Tips for better AI-generated images: what you can change in the prompt to get more out of any model.