---
title: "Creating images"
description: "Generate images on the Clip canvas: writing prompts, reference images, settings, style presets, choosing a model, and making several at once."
url: https://clip.sg/docs/creating-images
last_updated: 2026-09-16
---

# Creating images

Describe a picture and Clip makes it, on the canvas, ready to edit.

## The generator

Pick **Image Generator** from the tool rail. A node lands on the canvas and its form opens.

![The image generator](https://clip.sg/docs/images/image-generator.webp)

### Prompt

The only required field. Write what you want to see. Concrete beats clever: subject, setting, light, and the kind of photograph or illustration it should be.

> A matte black ceramic pour-over coffee dripper on warm linen, soft morning light from the left, steam rising, editorial product photography, shallow depth of field

### Reference images

Add up to fourteen of your own images to steer the look — a colour palette, a product you need to appear, a style you're matching. Upload them, pick one already on your canvas, or paste a URL.

References guide; they don't guarantee. For an exact object, put it in as a reference *and* describe it.

### Settings

**Size** runs from 1K to 4K depending on the model. Larger costs more and takes longer; 1K is right for most drafts and social work.

**Aspect ratio** offers ten shapes — square, portrait, landscape, story, banner and more.

### Style

**Add style** attaches a named visual style so the result matches a defined look. The node remembers it, so every generation from that node stays consistent. See [Style presets](https://clip.sg/docs/styles).

### Model

The model that makes the picture. Each has different strengths and a different price. Leave it on the default until you have a reason — the default is a strong general-purpose model. See [Choosing a model](https://clip.sg/docs/choosing-a-model).

## Generating

The button carries the exact cost — **Generate ⚡14** — and that figure updates as you change size, ratio and options. What's on the button is what you're charged.

![A generation in progress](https://clip.sg/docs/images/generating-node-on-canvas.webp)

The form is replaced by the running state, with the prompt and a way out:

![A generation running in the panel](https://clip.sg/docs/images/generation-in-progress.webp)

While it runs you get a progress bar, a label for the current stage, and an elapsed timer. You can:

- Carry on working elsewhere on the canvas.
- Close the panel, or the whole project.
- Reload the page — Clip picks the generation back up.
- **Cancel** it.

If your connection drops, Clip reconnects and recovers. A generation that's already running is not lost by a flaky network.

### If you're at your limit

Every plan allows a number of generations at once. Go past it and the extra ones are **queued** rather than refused — they show their place in line and start automatically as slots free. A queued generation can be cancelled for a full refund; once it's running, it's running.

## The result

The node becomes a real image, named from your prompt, placed on top of everything else.

![A finished image and its toolbar](https://clip.sg/docs/images/generated-image-on-canvas.webp)

Select it and the AI toolbar appears. From here you can change what's in it, cut out the background, extend it, upscale it, or split it into layers — see [Editing images](https://clip.sg/docs/editing-images).

To recreate or tweak a result later, right-click it and choose **Details** for the exact prompt and settings that produced it.

## Making several at once

Drop several generator nodes and run them together — they generate in parallel, up to your plan's limit. It's the fastest way to explore directions rather than iterating one at a time.

For many variations at once, describing the job to the [design chat](https://clip.sg/docs/design-chat) is usually quicker: it dispatches the batch and arranges the results for you.

---

**Next:** [Creating video](https://clip.sg/docs/creating-video).