ChatGPT Images 2.5 Adds Sketch: Draw a Doodle, Get an AI Image

OpenAI announced ChatGPT Images 2.5 on Tuesday, adding a feature called Sketch that lets users draw a rough doodle directly inside ChatGPT and use it as a reference for image generation. Type @Sketch into the chat box, draw something on the canvas that appears, then describe what you want. OpenAI's own test turned a mouse-drawn cat scribble into a realistic photo. The update also adds the ability to comment on specific parts of a generated image.

What happened

Detail	Fact
Product name	ChatGPT Images 2.5
Announced	Tuesday
New feature	Sketch
How to activate	Type @Sketch in the chat box
Additional change	Inline comments on parts of images

OpenAI shipped ChatGPT Images 2.5 on Tuesday with a new input method for image generation. The feature, called Sketch, opens a drawing canvas inside ChatGPT when you type @Sketch into the chat box. You draw something, add a text prompt describing what you want, and ChatGPT generates an image based on both inputs.

The idea is that a rough spatial layout, even a bad one, gives the model more to work with than text alone. According to The Verge, a mouse-drawn cat doodle was successfully converted into a realistic photo following simple instructions. No stylus or drawing tablet required.

The update also lets users leave comments directly on specific areas of a generated image, which should make iterative editing more precise than describing changes in plain text.

Why it matters

Sketch lowers the barrier for anyone who struggles to describe a composition in words. Designers, marketers, and e-commerce operators who need a quick mockup of a layout, product placement, or scene can now rough it out with a cursor instead of writing a long prompt.

The inline comment feature is arguably just as useful. Being able to point at a specific region of an image and say “change this part” is a more natural editing workflow than writing “the object in the lower left corner of the image.” It brings image editing closer to how people annotate documents or design files.

For teams already using ChatGPT for content production, this reduces the round-trips needed to get an image close to what they had in mind. That compounds over time when you’re producing a lot of visuals.

Our take

Sketch is a genuinely practical addition, not just a demo feature. The @Sketch trigger keeps it low-friction: it lives inside the existing chat interface rather than a separate tool. The real question is how well it handles vague or ambiguous sketches at scale, since a mouse doodle leaves a lot of interpretation to the model.

The inline commenting is the part we’d test first. Iterating on AI images has historically been clunky because text prompts are a poor substitute for pointing. If this works reliably, it cuts editing time in a meaningful way for teams doing regular visual content work, including ad creatives and social posts.

If you’re already exploring AI image tools for your business, this is a good moment to look at how AI integration fits into your content workflow. And if you want to see how that plays out in practice, our AI news coverage tracks these updates as they roll out.

What to do about it

1. Open ChatGPT and type @Sketch in any chat to activate the drawing canvas.
2. Sketch a rough layout of the image you want, then add a text description in the prompt field.
3. Generate the image and use the new inline comment tool to mark up specific areas you want changed.
4. Test it against your current image generation workflow to see if it reduces prompt iterations.

If your team produces ad creatives or product images regularly, Sketch and inline comments are worth a 30-minute test this week.

Source: The Verge · AI