On 28 October 2026 GitHub announced that Copilot can generate and edit images from a plain-English request. It works in Copilot chat in VS Code, the GitHub Copilot app and Copilot CLI. The chat model calls a dedicated image tool, GitHub names it as GPT Image 2.5 Flare, shows a preview inline and saves the file into the project.
What GitHub has confirmed
- Status: public preview for Copilot Pro, Pro+, Business and Enterprise. No general availability date is given.
- Billing: image generation is billed separately from the chat model, based on image tokens that count towards the plan’s AI credits.
- Controls: organisation and enterprise owners can switch it off by disabling the image generation model in Copilot model policies. An opt-in “Made with AI” watermark is available and is off by default.
What it means
Coding assistants are no longer only writing code. They now produce the placeholder screens, icons and marketing images that used to wait for a designer. For an early prototype that is a real saving in time. For a product you will put in front of customers, it raises two questions: who checks that generated assets are right for your brand, and who owns the result?
It also adds another line to the AI bill. Teams that adopted per-seat tools are finding that usage-based credits are what actually drives cost, and image generation draws on the same pool.
So what
If a supplier is building your app with AI tools, ask whether generated images are allowed in the project, whether they are labelled, and whether image usage is inside the quoted budget. Treat generated assets as drafts until a person has approved them. Our AI-assisted development service sets these rules out at the start, and you can start a project to discuss yours.