Skip to main content
The model is the image or video engine. The avatar is the person reference. They are separate choices: changing from one image model to another does not choose a different saved person.

​
Start with the selected default

Use the default offered by the tool for your first test. Keep the same source images, prompt, ratio, and image count while trying another model. Compare the things that matter for your work: product fidelity, face consistency, composition, and usable detail. The model picker only shows options available for that workflow and account. Generate, Virtual Try On, Editor, and Video can offer different models and settings.

​
Choose quality for the stage of work

Use a smaller available output while exploring a scene or pose. Once you have a promising setup, use the quality needed for delivery and inspect it at full size. Higher quality can cost more credits and take longer. Check the number next to the submit action after changing the model, quality, duration, or image count. Do not assume the previous cost still applies.

​
Spend on useful comparisons

Start with one image. If the garment is wrong because the source is unclear, a larger batch repeats the same uncertainty. Improve the source or brief, then generate a few candidates to compare natural variations. For a catalog, choose a setting that works across several representative products before running a larger series. Include a patterned item, a dark item, and a garment with details that are easy to lose.

​
Finish deliberately

Review the chosen result before upscaling. A larger file is helpful for delivery, but it does not replace checking the product against the original. For API integrations, use the actual identifiers in API models and settings; interface labels are not always API values.