More image options help when the team knows which detail to preserve or improve.
Figma’s November 18 release notes add two ways to explore generated images. A user can ask one model for two, four, or six images from a prompt. Alternatively, the same prompt can run across up to three models, producing one image from each.
The distinction is practical. Several outputs from one model explore variation within that model. Outputs from different models compare different interpretations of the brief. Mixing those experiments makes it harder to tell why a result improved.
For a founder producing product illustrations, campaign concepts, or interface assets, the new controls can shorten the path to a useful candidate. They also make it easier to spend credits on options that lack a clear review criterion. A short creative brief is still the starting point.
Choose between variation and model comparison
Figma’s image-generation help page describes the model picker and the multi-image options. It also explains how to preview an image in a selected layer or place the options on the canvas for comparison.
| What you want to learn | New generation choice | What to hold steady |
|---|---|---|
| Whether a model can produce a better variation | Two, four, or six outputs from one model | The prompt, reference, and selected model |
| How models interpret the same request | One output from each of up to three models | The prompt and reference material |
| Which candidate fits the real layout | Preview in place or compare on the canvas | The destination frame and visual requirements |
If a six-image set produces one good result, that shows a candidate can be found within those variations. It does not establish how consistently the model would reproduce the important details. Save the selected image and the settings that produced it.
A three-model comparison answers another question. The same brief may lead to different framing, object placement, or text treatment. Review those differences against the job rather than choosing the most decorative image in isolation.
Access still depends on the file and the product surface
The current help page lists the image feature across Figma products and requires edit access to the file. It says the tools are available on all plans, while the Starter plan’s AI image tools are limited to Figma Design.
That matters when a teammate can view the file but cannot use the generation controls. Check file permissions and the product surface before assuming that an absent action requires a different model. A plan’s broad eligibility does not mean every Figma surface offers the same workflow.
Figma also uses shared AI credits across its AI features and products. Image exploration therefore belongs in the team’s wider usage plan, alongside other AI work. The availability of a button does not establish that its use is unlimited.
Before a large exploration round, review the displayed generation options and the applicable credit information. Model selection and output count are part of that decision. Keep the actual usage record if the team wants to compare the expense of different creative approaches.
Write the constraints that make an option usable
A useful brief names the deliverable and the details that must survive. For a product illustration, that might include the number of objects, their relative position, required transparent space, and any text that must remain exact. For a campaign concept, it might include the subject, framing, and intended crop.
Separate those requirements from preferences. “Exactly one bottle with the approved label” is a requirement. “Warmer light” may be a preference. An image that satisfies the preference while changing the label still fails the product brief.
An existing reference image can make the target more concrete. Use material the team is entitled to supply, then check whether each output preserves the relevant details. A reference narrows the request; it does not establish that the output follows it correctly.
Our earlier report on Figma’s code layers and related controls covers another part of the design workflow. The image changes address asset exploration, so they deserve their own acceptance criteria instead of borrowing the checks used for working code.
Make the comparison small enough to interpret
A first round can hold the prompt and reference constant while generating a small set. Record which outputs meet the required constraints, then review the remaining candidates for composition and fit.
For example, suppose a six-output round yields two candidates that preserve the required object count and text. The usable share in that illustrative round is two divided by six, or about one third. That figure describes the selected round under its review rule; it is not a measured Figma success rate.
If a later round changes the prompt, model, reference image, and output count at once, the team may find a better picture without learning which change helped. Make one meaningful change at a time when the purpose is to improve repeatability.
The same principle applies to model comparison. Keep the review criteria constant across models, and inspect the images without relying on the model’s reputation. A preferred style is useful only if the candidate still meets the product’s constraints.
Count acceptable assets alongside credit use
The relevant production measure is often the expense of a usable asset, not the expense of one generated image. Keep the total credits used across rejected and accepted candidates, then divide by the number of assets the team can actually use.
Consider a hypothetical exploration costing 60 credits in total. If three outputs pass review, the average is 20 credits per accepted asset. If only one passes, it is 60. These figures illustrate the accounting method; they are not Figma’s published price for a generation.
Review time is another useful measure. A large set can save prompting time while adding selection work. Record the time spent checking text, shapes, object counts, and layout fit if those tasks dominate the production process.
A team can use that record to set a sensible budget for the next round. The aim is to understand whether another round is likely to produce a materially better asset, rather than continuing because it’s easy to request more options.
Keep the chosen asset connected to its brief
When an image is selected, retain the final brief, reference, model choice, and output version with the design file. That makes a later revision easier to review. Compare a new variant against the same requirements rather than someone’s memory of the previous image.
Inspect the actual export and crop too. An option that looks good on the canvas may lose important content at the size used in a product card or social post. The final destination is part of the asset’s job.
Figma’s update makes it easier to explore choices. A clear brief, a stable review rule, and an honest count of accepted assets turn that convenience into a production workflow the team can budget and repeat.



