GPT Image 2.5 prompt — turn a drawing into a photorealistic image

Preserve the layout, add the materials. The sketch decides where; the prompt decides what of.

A photorealistic render following the drawing's exact layout, proportions and perspective
The original rough drawing used as the structural reference
InputResult
Settings the Turn a rough drawing into a photograph of the same thing prompt ran at
ModeModelQualitySizeRatioCredits
Sketch to ImageThinkStandard1K2:310 cr

Published by OpenAI as an example for this model — not generated on this site. Model: gpt-image-2.5-sunburst. Source: OpenAI The published example ran at 1024x1536, a legal size we do not currently offer; the frame here is the same shape at our nearest size.

A rough drawing in, a photorealistic render out, with the composition and perspective untouched. The prompt divides the job cleanly in two: the sketch owns geometry — layout, proportion, perspective — and the prompt owns substance — materials, lighting, environment. The last line, banning new elements and text, is what stops the model treating your drawing as a suggestion.

In full

The prompt, nothing cut

Turn this drawing into a photorealistic image.
Preserve the exact layout, proportions, and perspective.
Choose realistic materials and lighting consistent with the sketch intent.
Do not add new elements or text.
211 characters

Run it here

This exact prompt, at these exact settings

The mode and all five settings are locked to this template, so there is nothing to line up by hand. Press Run this prompt above and the panel opens filled in — prompt, mode, quality, size and frame — and your first image needs no account. If your plan sits below what this ran at, the line under the panel says so.

PNG, JPEG or WebP · up to 25 MB

Words written on the sketch are read as instructions, not copied as text.

0 / 4000

No card, no waitlist. Credits only start after the free one.

This example was published at Think · Standard · 1K. A free run is Fast · Standard · 1K, so expect less fine detail and softer small text. The composition is the same; the finish is not.

3 levers

What is safe to change

Most libraries publish this list and stop, which is why so many copied prompts come back worse than the original: the swappable nouns are the safe half. The three sections under it are the other half — the clauses that are doing the work, the ones that break it if you touch them, and why it is written in this order.

  • The geometry clause

    "Preserve the exact layout, proportions, and perspective" is what makes this a render of your drawing rather than a picture inspired by it. Three properties, all of them things a generative model would otherwise happily improve.

  • "consistent with the sketch intent"

    The clause that hands material choice to the model while keeping it tethered. Replace it with your own material list when you know what you want — "brushed aluminium and walnut" — and the render stops guessing.

  • The addition ban

    "Do not add new elements or text" prevents the most common sketch failure: the model filling your empty regions with plausible furniture, foliage or signage because a photograph of that scene would have some.

Three ways to break it

Each one is a real failure with a reason attached. A rule without a reason is not usable.

  1. 01Dropping the addition ban

    A rough sketch has empty space, and a photograph of the same scene would not. Without the ban the model populates the gaps — and the things it adds are exactly the things you left out deliberately.

  2. 02Feeding it a sketch with handwritten labels and expecting them rendered

    This template says "do not add text". If your drawing has words on it describing what to draw, you want the sketch scene's own instruction instead, which reads handwriting as a brief rather than as lines to reproduce.

  3. 03Sketching too loosely and blaming the render

    The prompt promises to preserve the layout, and it will — including the parts that are ambiguous. A line that could be two different objects gets resolved by the model, and it does not tell you which reading it picked.

Why it is written this way

Inside the Turn a rough drawing into a photograph of the same thing prompt

Sketch-to-render is the workflow that most clearly separates what a person is good at from what a model is good at, and this prompt draws the line in exactly the right place.

The person is good at geometry. Where the horizon sits, how big the thing is relative to its surroundings, what is in front of what, which way the eye travels. Those are decisions, and they are hard to specify in prose — a paragraph describing a composition is far less precise than thirty seconds of drawing it. The model is good at substance: what brushed metal looks like at that angle, how afternoon light falls across a surface, what the ground would be doing under all of it.

So the prompt hands each half to whoever does it better. "Preserve the exact layout, proportions, and perspective" gives geometry to the sketch. "Choose realistic materials and lighting consistent with the sketch intent" gives substance to the model. Four lines, and the division of labour is complete.

The last line is the one people leave out and then complain about. "Do not add new elements or text." A rough sketch is sparse by nature — you drew the thing you cared about and left the rest blank. But a photorealistic image of that scene would not be sparse; it would have the clutter that real environments have. So the model helpfully supplies it, and you get your object surrounded by plants, furniture and signage you never drew. The ban is one clause and it is the difference between a render of your idea and a render of the average scene containing your idea.

The clause worth adapting rather than copying is the material one. "Consistent with the sketch intent" is a delegation, and it is the right default when you are exploring. When you already know the answer, replace it with the specifics — a named material list constrains the render far more tightly and removes the main source of surprise.

The limitation is honest and worth stating: the prompt preserves your layout including its ambiguities. A line that could read two ways will be resolved silently, and the resolution will look confident. If a shape matters, draw it twice as clearly as feels necessary.

4 questions

Turn a rough drawing into a photograph of the same thing — common questions

How rough can the drawing be?
Rough is fine; ambiguous is not. The prompt preserves your layout including the parts that could read two ways, and it resolves them silently.
Why did it add things I did not draw?
Missing addition ban. A real photograph of your scene would have clutter, so the model supplies it. "Do not add new elements or text" prevents that.
Can I control the materials?
Yes, and you should when you know. Replace "consistent with the sketch intent" with a specific material list and the render stops guessing.
What if my sketch has handwritten notes on it?
Use the sketch scene rather than this prompt. Its instruction reads handwriting as a description of what to draw instead of as lines to reproduce.

Written and maintained by Andy SwiftPublished Last updated

Your first GPT Image 2.5 image is free

Generate my first image