How-to · 10 Nov
How-to · 3 min read
How to keep one consistent style across a set of AI images
You are making a series of images with AI, and every new generation drifts a little further from the look you started with. It might be character sheets, a game's item icons or illustrations for one client. The colours shift, the line quality changes, and the character's face is never quite the same twice. You want the tenth image to look like it belongs next to the first one, without rebuilding your prompt and settings from memory every session.
- Published
- 17 Nov 2026
- Delivery
- No. 12
- Topic
- Visual and media
- Evidence
- 8 claims quoted from abrush.ai
01
Why it's hard
Most AI image services treat each generation as a fresh start, so the only thing tying a series together is a prompt you keep copying around. Small changes in wording, model version or seed are enough to change the style, and web tools rarely let you save the full set of settings behind a result you liked. Text alone is also a weak way to describe a style; a reference picture or a trained style model says far more than adjectives do. And when the images end up in Photoshop anyway, every hop between a browser tab and your document is another chance to lose track of what produced what.
This part describes the usual way. It makes no claims about ABrush, so nothing is quoted here.
02
How ABrush does it
ABrush is a plugin that runs AI generation inside Adobe Photoshop, and several of its controls are about repeating a look. Generation history keeps your past results, and presets save the prompt together with the settings, so you can return to the exact recipe behind an image you liked. For style, it supports LoRA style models and IP-Adapter references, and ControlNet for guiding composition. Because everything happens in your document, you can fix small differences by hand before moving to the next piece. It works with more than 23 models, so you can try a few early on and then commit to one for the series.
Save your generation history, use presets (prompts and settings), and return to your best results without starting from scratch.Quoted from abrush.ai
Adjust details, match your style, and stay in control of every output without ever leaving your workspace.Quoted from abrush.ai
Presets and preset sharingQuoted from abrush.ai
03
Step by step
- 1Generate a first image you are happy with, trying a few of the available models until one fits the look.
- 2Save the prompt and settings as a preset so the recipe is stored, not just the picture.
- 3Add a reference image or a LoRA style if the look depends on a specific palette or drawing manner.
- 4Generate the next pieces from the same preset, changing only the subject in the prompt.
- 5Compare results in the history, then match small details by hand in Photoshop.
04
Tips
Pick one model for the whole series. Different models interpret the same prompt in their own way, so switching between them mid-project is one of the quickest ways to lose a consistent look.
Work with different models (Stable Diffusion, Flux, Gwen Image and others) and choose the best option for each stage of your workflow.Quoted from abrush.ai
Share presets with the rest of the team. If two artists work on the same project, a shared preset keeps both of them generating from the same settings instead of two slightly different versions of the prompt.
Presets and preset sharingQuoted from abrush.ai
Keep an eye on what the plugin changes in your text. When ABrush translates or improves a prompt, the result is always shown in the text field, so you can catch a rewrite that would push the style somewhere else.
Whether it's translation or prompt improvement, the result is always shown in the text field.Quoted from abrush.ai
Use a reference image for anything hard to put into words, like a brush texture or a lighting mood. One good example picture usually carries a style more reliably than a long list of adjectives in the prompt. Keep that picture with the preset so the next session starts from the same pair.
General advice. Not a claim about the product.
05
Questions
Can I use my own style models?
Yes. The plugin lists LoRA styles among its pro controls.
LoRA stylesQuoted from abrush.ai
Does it support reference images for style?
Yes. IP-Adapter and references are part of the pro controls, alongside ControlNet.
IP-Adapter/ReferencesQuoted from abrush.ai
Who owns the images I generate?
You do. ABrush states that you own every image you generate and that it does not train on your data.
You own every image you generateQuoted from abrush.ai
More from ABrush
All 3 answersHow-to · 6 Nov