Generating images

Image mode turns a prompt into pictures you can save to your media and use in designs. It runs on Qwen, OpenAI, or Google, the three providers that generate images natively, or on BytePlus. See the provider matrix for details.

Quick overview

1. Switch the composer to Resim (image) mode.
2. Write a prompt and set the aspect ratio, resolution, and count.
3. Optionally pick a Photo Studio mode and add reference images.
4. Generate, review in the built-in editor, and save to your library.

Controls

Control	Options
Aspect ratio	1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9
Resolution	1K or 2K
Count	1 to 4 images per generation

Photo Studio modes

Photo Studio tunes the generation for a kind of shot, so you get a fitting look without hand-writing every camera and lighting detail. Each mode exposes its own set of Camera, Direction, and Style controls, such as lens, angle, composition, lighting, and palette.

Mode	Best for
Product Photography	Clean, catalog-style product shots
Portrait	People and headshots
Fashion Photography	Styled fashion and apparel looks
Close-up / Macro	Detail, texture, and small subjects
Cinematic Shot	Filmic, dramatic scenes
Illustration & Art	Non-photographic looks: painting, 3D render, anime, and similar styles, without camera controls

Every photographic mode groups its options the same way: Camera covers things like camera body, lens, focal length, and aperture; Direction covers angle, composition, crop, pose, and gaze; Style covers lighting, background, and color palette. Illustration & Art keeps the Style controls but drops the camera-specific ones, since there's no literal camera in an illustrated shot.

To give you a sense of what the controls hold, Camera exposes options such as a camera body or format (for example Studio Digital, Medium Format, a Phase One technical look, Full-Frame DSLR, Film Analog, or a raw handheld look), a Lens type, a Focal Length (24, 35, 50, 85, or 135mm), and an Aperture (roughly f/2.8 through f/16). Direction exposes an Angle or Shot such as Eye Level, a top-down Flat Lay, Three-Quarter, Low Hero, or Macro Detail, and a Composition such as Rule of Thirds, Negative Space, Symmetry, Foreground Frame, or Center Hero. Style exposes a Lighting such as Softbox, High Key, Dramatic Rim, Gradient Glow, or Hard Flash, and a Background or Surface such as Seamless White, Polished Stone, or Acrylic Reflection. These are examples, not the exhaustive list.

The Photo Studio kit also does motion

The same kit that powers these still modes also offers a motion side that turns a product shot into a short clip, with modes such as Product Motion or a full Product Showcase Reel. That handoff runs through video generation, see the Photo Studio kit for how the kit is organized.

Reference images

Add reference images and tag each one with a role, so the model knows what it's looking at rather than guessing. Tagging accurately steers the result toward the subject, style, or composition you already have in mind, rather than blending everything together. The role set changes with context.

Product roles	Fashion roles
Product	Dress
Logo	Top
Packaging / label	Bottom / pants
Mockup / template	Jacket / outerwear
Environment / set	Hat
Surface	Bag
Prop	Shoes
Material / texture	Jewelry
Color palette	Sunglasses
Style reference	Accessory
Other	Model, fabric, location, style reference, other

Step by step

Open AI Studio from the app sidebar (the item tooltipped Create with AI), then switch the composer's mode control to Resim (Image). Pick an image-capable provider (Qwen, OpenAI, Google, or BytePlus). If none is connected, add one from Settings, AI Settings, Provider & Model tab (see Connecting a provider).

1. A first plain-prompt image

1. Write a prompt, for example a matte black ceramic coffee mug on a light oak table, soft morning light.
2. Set the aspect ratio (for example 1:1), the resolution (start at 1K), and the count (for example 2).
3. Click generate.
4. Review the results in the built-in editor. Each attempt stays in the thread, so you can scroll back and compare.
5. Save the one you like to your media library, from where it drops into any design the same way an upload would.

2. A Photo Studio product photo

1. Open Photo Studio and pick Product Photography.
2. Under Camera, choose a body such as Medium Format, a Focal Length of 85mm, and an Aperture around f/8.
3. Under Direction, set the Angle to Eye Level and the Composition to Rule of Thirds.
4. Under Style, set Lighting to Softbox and the Background to Seamless White.
5. Add a reference image of your product and tag it with the Product role. Add a fabric or finish swatch and tag it Material / texture, or a palette chip tagged Color palette.
6. Write a short prompt describing the scene, then generate.

3. Let auto-configure set it up for you

1. Type a plain description, for example a matte black water bottle on a marble counter.
2. Optionally attach a reference image on a vision-capable provider.
3. Click the wand button. It sends your prompt (and the attached image) to your active chat provider, which picks a fitting mode, for example Product Photography, and fills in a matching camera and lighting combination.
4. If no chat provider is connected, a local keyword match sets reasonable defaults instead (see Troubleshooting).
5. Adjust anything you want, then generate.

4. Fix one part with a region edit

1. Open a generated result in the built-in editor.
2. Choose the masked region edit, then paint over just the part you want to change, for example a label or a background corner.
3. Describe the change and apply it. Only the masked area regenerates, the rest of the image stays exactly as it was.
4. Use crop and rotate in the same editor for framing, then save the result to your library or share it to the styles gallery.

Refining

A wand button runs an auto-configure pass: it sends your prompt, and an attached image on a vision-capable provider, to your active chat provider, which picks a fitting mode and control values for you, so you can start from a stronger baseline than typing every setting by hand. If no chat provider is connected, a local keyword match sets reasonable defaults instead.

Once you have a result, the built-in editor lets you crop, rotate, and make a masked, AI-driven region edit to touch up part of the image without regenerating the whole thing. From there, save the result to your library, or share it to the styles gallery.

Common tasks

Concrete prompts to copy and adapt. Pair each with a matching Photo Studio mode for a stronger result.

Goal	Example prompt	Mode to pair
Catalog product shot	a frosted glass perfume bottle on polished stone, gradient glow lighting, plenty of negative space on the right for text.	Product Photography
Lifestyle flat lay	top-down flat lay of a skincare set on a white marble surface, minimal props, high key lighting.	Product Photography, Flat Lay angle
Headshot	editorial headshot of a smiling barista in an apron, soft window light, shallow depth of field.	Portrait
Texture detail	extreme close-up of woven linen fabric, raking side light to show the weave.	Close-up / Macro

For a catalog shot with a real product, add the product photo as a tagged Product reference rather than describing it in words, it steers the result far more reliably.

Where results go

Every image you generate is kept in the chat thread it came from, so you can scroll back and compare earlier attempts at the same prompt. Saving a result adds it to your media library, from where it can be dropped into any design the same way an upload would be, and sharing it to the styles gallery turns the prompt and settings that produced it into a reusable recipe for others.

Iterate cheaply

Generate at 1K with a small count first to find a direction, then raise the resolution once you like the look.

Reference images beat long descriptions

A tagged reference image, a product shot or a swatch of fabric, usually steers a result more reliably than describing the same thing in words.

Troubleshooting

Auto-configure just picked generic defaults. The smarter auto-configure pass needs a connected chat provider to read your prompt and set a fitting mode. With no chat provider connected, the wand falls back to a local keyword match, which sets reasonable but plainer defaults. Connect a chat provider in Settings, AI Settings, Provider & Model tab to get the smarter configuration.

My reference image did not steer the result. Tag each reference with a role (Product, Material / texture, Color palette, Style reference, and so on) so the model knows what it is. Untagged references tend to get blended together instead of guiding the subject, style, or composition you had in mind.

I only want to fix one part of an otherwise good image. Use the masked region edit in the built-in editor instead of regenerating the whole thing. Mask just the area to change, describe the change, and apply, so the rest of the image stays put.

Image mode has no provider to run on. Image generation runs on Qwen, OpenAI, Google, or BytePlus only. If image mode looks empty, connect one of those from Settings, AI Settings, Provider & Model tab. See the provider matrix for which one covers what.

Related

• Connecting a provider
• Provider and model matrix
• Skills and kits for the Photo Studio kit that powers these modes
• Styles gallery