Create and use AI Styles

An AI Style is a reusable reference set for visual direction. It helps related image edits begin from the same source material instead of reconstructing a look from memory or a long prompt.

AI Styles guide generation. They do not lock exact color, lighting, composition, or product fidelity. Review every result against its source render.

Build a coherent reference set

1. Open Library and choose AI Styles.
2. Create a style.
3. Give it a name another person can recognize.
4. Add the supported reference images.
5. Review the set as one visual direction.
6. Save the style.

Choose references that agree on the qualities you want Glossi to infer. A smaller coherent set is more useful than a large mood board containing several competing art directions.

Name what the style controls

Use a name tied to observable direction:

• Warm gallery / low contrast
• Spring launch / cobalt and chrome
• Soft daylight / pale concrete

Avoid labels such as Premium or Modern. They do not tell the next person what the style is meant to preserve.

Apply a style to a render

1. Open a completed image in Renders.
2. Start AI Edit.
3. Choose the AI Style.
4. Add a short instruction only when the result needs product- or campaign-specific context.
5. Generate candidates.
6. Compare each result with the source render.

Inspect product shape, labels, material boundaries, small hardware, and placement before accepting the derivative. When those details must be exact, make the change in Studio instead.

Edit a render with AI

Use a style in an automation

A supported Style node can create related derivatives after a render stage. Test the style on representative products before placing it in unattended work.

One accepted result does not prove that the style will preserve every product category. Monitor automation runs and review the generated versions according to the risk of the destination.

Build an automation

Maintain the style

• Remove references that introduce an unintended treatment.
• Keep separate campaigns or product categories in separate styles.
• Rename a style when its reference set changes meaningfully.
• Preserve accepted outputs as evidence of what the style produces in practice.

The value is shared visual direction with a visible source, not a promise that generative output will be identical every time.