EditorNodesPricingBlog

DRAW TO EDIT

Sketch it, and let the model finish it

Some things are far quicker to draw than to describe. Rough out where everything sits on the built in canvas, write a prompt covering what those shapes are, and get a finished image back.

Open the canvas

The problem with describing a composition

Text is a poor tool for spatial arrangement. Getting a model to place a figure slightly left of centre, with a doorway behind their right shoulder and the horizon low in frame, takes several attempts and often never quite lands. Every ambiguity in the sentence gets resolved by the model rather than by you.

Thirty seconds of rough shapes settles all of it. The sketch carries placement, proportion and depth relationships directly, and none of that has to survive a round trip through language.

That leaves the prompt free to do what prompts are good at, meaning material, lighting, mood and treatment. The division of labour is the whole point.

The Draw to Edit canvas

One surface holding the drawing tools, the prompt and the model settings together.

Draw to Edit canvas with drawing tools alongside the prompt and model controls

The Draw to Edit canvas. Sketch on the left, prompt and model settings alongside it.

1

Sketch, prompt, generate

Draw to Edit gives you a canvas inside the workspace. You sketch the arrangement you have in mind, add a prompt describing what it should become, pick a model, and the system turns the drawing into a finished image.

The drawing does not need to be good. It needs to be clear about placement, since that is the information a prompt struggles to carry. Where the horizon sits, which side the figure stands on, how large the building is against the sky.

Behind the scenes this runs as an image-to-image task, with your sketch as the source. That is worth knowing because everything true of image-to-image prompting applies here too.

Settings reference

What Draw to Edit takes as input and what each part contributes.

Canvas drawing

Your sketch, which acts as the source image. Composition and placement are what it contributes, not finish quality.

Drawn in the browser

Prompt

Describes what the sketch should become. Covers subject, material, lighting and treatment, since the drawing already fixes the arrangement.

Free text

Model

The editing model that interprets the sketch. Processed as an image-to-image task, so the editing roster applies.

Image to image models

Sketching for a model rather than for a person

A sketch for another human can be loose, because they fill in intent from context. A sketch for a model should be clear about boundaries, since that is what it reads. Blocked in shapes with definite edges work better than fine lines suggesting a form.

Relative size carries meaning. A figure drawn small against a large shape reads as distance, and the model will honour that. Getting proportion roughly right in the sketch saves arguing about scale in the prompt.

Then let the prompt take over. Once the sketch has settled where things are, describing the layout again in words competes with the drawing instead of supporting it.

Related tutorials

Coverage of the editing models that interpret your sketch.

Nano Banana Pro 2 image-to-image

How the editing models read a source image, which is what a sketch becomes.

FLUX.2 image generation guide

Detail handling across the FLUX.2 family when interpreting an input image.

Draw to Edit questions

Do I need to be able to draw?

No. The sketch communicates placement and proportion, not craft. Rough shapes marking where things sit relative to each other carry the information the model needs, and that is a very low bar to clear.

Composition. Describing a precise arrangement in words is slow and unreliable, and the model resolves ambiguity however it likes. A sketch fixes the arrangement directly, so the prompt only has to describe what things are rather than where they go.

It runs as an image-to-image task with your drawing as the source. The practical difference is that you produce the source in the browser instead of supplying one, which makes it the fastest route when the picture you want does not exist yet.

Describe what the shapes are and how they should look. The drawing has already settled placement, so spend the prompt on material, lighting, time of day and treatment. Restating the layout in words competes with the sketch.

When the arrangement you want does not exist to photograph. If a reference image already shows the composition, use it directly through image-to-image. Draw to Edit is for the case where the layout is only in your head.

Yes. The output is a completed task like any other, so it can be picked as the source for a further edit, isolated with background removal, or upscaled for delivery.

Other ways to generate images

Draw to Edit is one of seven generation types in the image workspace.

Text to Image

Write a prompt and generate images from scratch. Pick the model, aspect ratio and output format, and add a LoRA style if you have one.

Image to Image

Start from an image you already have and describe the change. Useful for style transfer, edits and mood changes on an existing frame.

Character Generation

Upload one to three reference photos and keep the same character consistent across many generations.

Image Upscaling

Increase resolution on an existing image while keeping identity and detail intact, with a selectable upscale factor.

Background Remover

Cut the background out of an image with edge detection precise enough for compositing work.

Model Training

Train a custom FLUX LoRA on your own images so you can generate in your own style or with a specific subject.

    © 2026 - AI FILMS LLC