Generation settings
Generation settings shape what comes back: the shape of the frame, how many results you get, and how closely the output follows your words. You rarely have to touch them. The Auto model picker chooses a model for you, and every setting starts on a sensible default. Open them when you want more control.
Where the settings live
Section titled “Where the settings live”A few settings sit right in the composer, above the Generate button:
- Aspect Ratio — the shape of the frame.
- Images — how many results to make in one run (1 to 4).
Everything else lives behind Advanced settings, a panel that stays collapsed until you open it. The label on the right shows the model the settings apply to, or Select a model if you haven’t picked one.
The panel adapts to the model you’re using. A model that doesn’t support a setting won’t show it, so you only see controls that do something. See Choosing a model for how the Auto picker works and how to pin a specific model.
Aspect Ratio
Section titled “Aspect Ratio”The shape of the frame — square, landscape, or portrait. Pick a tile to set it. The options change with the model, but the common ones are:
| Ratio | Use it for |
|---|---|
| 1:1 | Social posts, profile images |
| 16:9 | Landscape, slides, banners |
| 9:16 | Stories, Reels, mobile screens |
| 4:3 | Traditional photo, print |
| 3:4 | Portrait photography, posters |
Some models offer more, like 2:1 or 3:2. The tiles show a small preview of each shape so you can scan them quickly.
Images
Section titled “Images”The Images stepper sets how many results to make in one run, from 1 to 4. Use the + and - buttons to change the count. More images mean more options to choose from in a single run.
The Generate button updates to match — it reads Generate 3 images when the count is above one.
Some models apply a built-in look to the output — realistic, illustration, and others, depending on the model. When the model supports styles, a Style menu appears in Advanced settings. Choose one, or leave it unset to let the model decide.
A Seed is a number that fixes the random part of generation. Leave it blank for a fresh result every time. Set a seed to get repeatable output:
- The same seed with the same prompt and settings gives you the same result again.
- Keep the seed fixed and change the prompt slightly to iterate on one composition.
The field reads Random when it’s empty.
Guidance
Section titled “Guidance”Guidance controls how closely the output follows your prompt. It shows on models that support it, as a slider in Advanced settings.
- Lower values give the model more room to interpret — looser, more creative.
- Higher values stick more literally to what you wrote.
The default is tuned for the model, so move it only when a result feels too loose or too rigid.
Negative Prompt
Section titled “Negative Prompt”A Negative Prompt lists what you want to leave out. The model tries to avoid anything you name here. The field reads What to avoid in the generation… and appears on the models that support it.
Use it for cleanup, like:
blurry, low quality, watermark— general quality controltext, letters, words— keep out unwanted text
Resolution
Section titled “Resolution”On models that support it, Resolution sets the output size. Higher resolution means more detail and a longer wait. Pick an option to set it; click it again to clear it and fall back to the model’s default. The choices depend on the model.
Mature content
Section titled “Mature content”If your Studio admin has enabled mature content, you see an optional toggle.
You rarely need any of this
Section titled “You rarely need any of this”The defaults are chosen to give good results without setup. Write your prompt, leave the model on Auto, and press Generate. Reach for these settings when you want a specific frame shape, a repeatable result, or tighter control over the output. For a walkthrough of the full create flow, see Creating images.