They say a picture is worth a thousand words, and with the new ChatGPT Images 2.5 model, that certainly rings true. Rather than spending ages crafting long, complex text prompts to describe what you want, you can draw a rough outline directly in ChatGPT to set the scene, then refine the details as you go.
You can use this feature with a free account, and your sketches can be as basic or as elaborate as you wish. While some outputs might carry that AI polish, the results are impressively accurate.
Most importantly, the process gives you far greater control over the end result because you can make tiny adjustments to any element of the image.
1. Activate Sketch
To use the Sketch feature, select Images from the sidebar (or on a mobile device, via the menu which is activated via the icon in the top-left corner) and choose Sketch.
Alternatively, you can type @Sketch in the text field and choose the Sketch option when it appears.
2. Start drawing
Now you can start drawing. You may find this is easier on a mobile device because you can use your finger or a stylus on the touchscreen. But you can also perform the task on a computer using a mouse or trackpad. You’re not looking to create a work of art, just something that resembles a picture.
3. Use the tools
As you draw, you can make good use of the tools available to you. At the top of the screen, you’ll see options to draw lines, enter text, use shapes and arrows and erase parts of your sketch. There are also undo and redo buttons.
To the left of the screen is a slider that determines the size of your brush. At the bottom of the screen is a set of colours along with a wheel that allows you to select any shade you wish.
4. Click the tick
When you’ve finished drawing, simply click the blue tick button. ChatGPT doesn’t just take the image and instantly convert it into a photograph, though. It asks you to describe how you want the image to look, and it will suggest a prompt.
You could accept the suggested prompt if you wish, or enter whatever you want. In my case, I explicitly asked it to create a photograph. When you're ready to generate, click send again.
5. Use the image
After a short while, ChatGPT will present the result. You can select the Share button to copy the link, send the picture to social media or download it.
6. Edit the image
If you’re not entirely happy with the image, you can select Edit and use the available tools.
You can mark up the image, select specific areas to add targeted feedback, remove the background, erase unwanted elements, or reframe the aspect ratio — choosing from square, portrait, story, landscape, or widescreen to fit your needs.
7. Action the changes
If you comment on a part of the image and click Send, ChatGPT will take action it and make the necessary amends. It’s a great way to ensure that every part of the image is to your liking and puts you in greater control.