Generating images
When agentic mode is on, generating an image is just part of the conversation — ask for one in plain language and it appears inline, ready to use in the rest of your work.
Asking for an image
Section titled “Asking for an image”Describe what you want:
Generate a wide banner image of a calm desk with a laptop and a cup of coffee, soft morning light, muted colors.
The model calls the image tool, generates the picture, and shows it in the reply. You can keep iterating in the same breath — “make it warmer,” “remove the coffee” — and refer back to earlier images as the conversation continues. Generated images can also flow into other steps: into an artifact like an HTML gallery, or into a workflow that produces a batch of them.
Choosing an image provider
Section titled “Choosing an image provider”Catalyst can generate images through more than one provider, and you pick which one in your settings:
- A built-in generator, available to everyone at no extra cost — the default.
- Premium providers (such as OpenAI or Gemini image models) for higher quality, where your plan includes access.
The conversation is the same whichever you choose; only the engine producing the picture changes — the same swap-the-engine principle as switching chat models.
Editing images
Section titled “Editing images”You can also hand the model an existing image and ask it to change it — the image-editing tool works the same way as generation, inline in the conversation.