The whole point of Prompt Studio is that the prompt and the output belong together. Until now you could attach images after the fact. Now you can generate them from the prompt itself.
Generate stays on the prompt
Generate is an action on a prompt you’re already looking at, in the Generations section of the editor. Not a separate canvas that forgets the phrase. You pick a model (Grok Imagine is one of them), and when it finishes, the image is pinned as an output next to the text that produced it.
That sounds small until you’ve lost the good version of a visual prompt because the image lived in one tab and the text lived in chat history. Same library, same prompt, same search later.
Import from more places
X and Reddit still work for prompts worth keeping. You can now also paste:
- share links from ChatGPT, Grok, Gemini, Claude, DeepSeek and Midjourney, which become prompt threads where possible instead of one dumped blob of text
- public image URLs, as references
- any page, which unfurls into a title, description and image
A calmer editor
The editor is read-only until you tap the pencil, and Done when you’re finished. Cmd/Ctrl+C copies the full phrase, and text-only prompts show snippets as thumbnails again.
More on the updates page and the project page. If you already have a prompt that makes visuals, open it and hit Generate once.