Guide

Choose settings with a small, fair test

Compare one setting at a time using the same brief and references, so you can choose what the job needs.

Max McIsted avatar

Max McIsted

18 November 2026

A hand holding a SOREN+KOLL Tomato Leaf Hand Balm tube in soft window light.

Let a small test inform the settings

A small comparison can help you choose settings for the work in front of you. In this example, the brief and product reference stay the same while the setting changes, making it easier to see what happens to skin texture, the grip and the label.

Before you run the test, decide which of those details the finished image needs to get right. You can then judge both results against that same standard and choose a setting that fits the job and budget, with something more useful to go on than a general model ranking.

A choice of settings you can explain through the brief, review criteria and cost of the job.

Before you start

  • Have one product brief and the reference images you’re allowed to use ready for both runs. Keeping the inputs identical makes it easier to see what difference the setting makes.
  • Decide how many credits you want to spend on the comparison, using the estimates shown in your current workflow. Start with one image for each setting and check the cost before generating.

Step by step in the app

In the image tool, check the model, output size, format, image count and cost. Make the first test with one image, then change only the setting you’re comparing, such as the available quality level. Keep the brief and references unchanged.

If SuperAgent is preparing the work, use the same principle when reviewing its plan: confirm what differs between the tests before approving. Compare both outputs at the same zoom and against the same product reference. Choose the least expensive setting that meets this job’s needs; the result of one test is not a universal ranking for every product or image style.

Decide what good looks like

Before you open the settings, decide what would make the image useful for this job. For the hand shot in this guide, that means skin that looks photographed, a believable grip and a label you can read. Keep those three checks beside your brief as you work through the test, so you judge both results on the same things rather than being drawn to whichever image looks most striking at first glance.

Check the available settings

Open the image tool and look through the model, output size, format and image-count settings available to you. The cost beside Generate updates as you make changes, so you can check what each test will cost before running it. Start with one image for each setting you want to compare. That gives you something concrete to review without committing credits to a larger run before you know what works.

The image generation panel with model, output and image-count controls above the Generate button and its credit estimate.

Change one thing

Use the same brief for both runs and change only the setting you want to compare, keeping the product photo, wording, format and everything else as consistent as possible. In this example, the first test uses Low and the second uses High. It can be tempting to improve the prompt between runs, but save that for a separate test; otherwise, you won’t know which change made the difference.

Compare at the same zoom

Once both images are ready, put them side by side at the same zoom and work through the three checks you set at the start. Look at the skin texture, follow how the fingers hold the product and read the label’s small print against your reference. Then view each image at the size it will actually be used, because detail that looks good enlarged still needs to be clear in the finished placement.

A comparison slider showing the same hand balm shot at the High and Low settings, at the same zoom.

Choose what the job needs

Choose the lowest setting that meets the brief and passes your checks. If neither result does, look for what they both miss before spending more: perhaps the product reference is unclear or the action needs a more precise description. Keep a note of the setting, references and checks behind your final choice, so the team can use what you’ve learned on similar work while recognising when a new product or placement needs another test.

A useful detail

Keep the inputs matched so you can judge what changed. Two attractive images made from different briefs won’t tell you which setting suits this job.

Review the work

  • Compare outputs made with the same brief and references. If another input changed, treat that result as a separate experiment.
  • Check product detail and composition at the size you will actually use. Inspect the label and grip closely, then step back to the intended display size.
  • Choose the setting that meets the brief within your budget. Record the choice for this job without turning one comparison into a general model ranking.

If it needs work

If neither result meets the brief, look for the problem they share before spending more. You may need a clearer reference or a more specific instruction; if those inputs are already clear and the fault keeps repeating, ask for help rather than choosing a winner simply because one image looks more polished.

Build on the result

Record the setting, references and checks that informed your choice. Reassess when the product or placement changes; a successful test is evidence for this job, not every future one.