Image Generation

Generate an image from a text prompt, with optional reference images. Results land on the canvas at the click position.

Open the panel

Click Generate Image in the bottom toolbar.

The panel anchors at the cursor; a generated result will appear where the panel sits.

Pick a model

The model picker shows every image model unlocked by your configured keys:

• WaveSpeed – Best, all models (Get API Key)
• GeminiGen — low-cost, but can be slower (good for Nano Banana)

The panel shows a short label per model. Click the model dropdown to switch.

Set the aspect ratio

Standard presets: 1:1, 3:4, 4:3, 16:9, 9:16, plus Custom (type pixel width and height).

The selected aspect ratio is constrained to what the model supports — the picker hides unsupported combinations.

Add reference images

Click the + in the Input Image (optional) slot to attach a reference. Useful for:

• Image-to-image — the model conditions on the reference. Strength is exposed in Advanced Settings.
• Style transfer — for models like FLUX Kontext.
• Subject preservation — for models with identity-preserving variants.

Multiple references can be added when the model supports them.

Write a prompt

Type into the prompt textbox. To inject media from the board, type @ and pick the element from the popover — it becomes part of the prompt context.

For voice input, click the microphone icon — see Voice Input.

Enhance a prompt

Click Prompt Enhancer. The enhancer analyses the attached reference image (if any) and rewrites the prompt with camera angle, lighting, environment details, and other concrete cues. See Prompt Enhancer.

Generate

Click Generate. The cost estimate is shown next to the button (e.g. ~$0.012). When the result returns it appears on the canvas next to the panel and is added to the generation history.

Iterate

After a result:

• Regenerate — same prompt, same model, new seed.
• Iterate — opens the panel again with the result attached as a reference for the next step.
• Refine — turns the output into an editable HTML card when it works better as live HTML you can keep tweaking.

Where to go next

• Prompt Enhancer
• Video Generation with Kling
• Mentioning Canvas Media
• Generation History