Image Generation
Last updated:
Image Generation makes a new image from your words. Start from a prompt alone, or add reference images to guide the subject or the style.

Charges apply to completed actions. Nova and Ultra reserve the standard amount and settle the actual charge when the action finishes, which can be lower.
How to generate an image
- Open Editor → Trends on desktop, or Create → Trends on mobile web.
- Type what you want in the prompt box. Add reference images with + if you have them.
- Pick a model and send. Kira opens a new project with the result on the canvas.
- Inside a project, choose Image Generation in the AI tools to open the full panel: prompt, references and aspect ratio. Choose Confirm to run it.
- Compare results in the version strip and Export the one you want.

Write a prompt that works
Name the subject, the setting and the look:
A cosy coffee shop on a rainy night, warm lighting, reflections on the window, cinematic photo.
Add the framing if it matters: "square product photo" or "vertical portrait". When you attach a reference, say what it is for, such as "use image 1 for the colour palette only". A reference guides the result; it is not copied pixel by pixel.
When the result misses something
- Shorten the prompt and put the important element first.
- Replace a blurry or busy reference with a clean one.
- Keep working in the same project and ask for one change at a time. Use Image Edit to change what you have instead of describing a whole new scene.
Want to make it move? Video Generation can start from this image. To reframe it, use Crop or Expand.
Every completed generation uses credits, and a running task reserves them first. See Credits & Usage.