How to select the Canvas

To select Canvas, open a new chat and select the model GPT-4o with Canvas in the model selection.

In this mode, you can use the canvas.

The model decides automatically when to use the canvas and when not (like with other tools like web search). If it does not automatically use the canvas, you can lead it a bit by asking to use this functionality. E.g. “use the canvas for this request”

How to use Canvas

Canvas opens automatically on the right after the prompt has been sent, and the answer is generated. There are different options to continue now:

In-line editing

You can edit the output in-line directly in the editor on the right. You can type and change the text or mark a section, format text or mark sections to re-write them and optimize these sentences only.

The formatting options are not shown for coding, as they are not needed. Instead, you can ask to describe or explain a marked section.

Optimizing the whole text generally

At the top, you will find different buttons to optimize the general characteristics of the text, like the length and style, and improve the text.

For code, you can add comments, fix bugs, change the language or add logs.

Editing in the regular chat

You can still give instructions on what to change in the regular chat if none of these options suit your situation.

Versioning

On the top right, the different versions are selectable if you want to return to a previous version of your text. When you are ready, you can use the copy button on the top right to copy the whole text into your clipboard and paste it elsewhere.

Canvas works best with GPT-4o at the moment. We will support Claude 3.5 Sonnet soon.