Happy Sunday evening, everyone. I used to leave my code editor every time I needed a picture for a project, and that small detour always broke my train of thought.
GitHub shipped image generation inside Copilot chat on October 28. You describe a picture in plain words, Copilot previews it in the same chat, and it saves straight to your project. It is in public preview, and it runs on a dedicated image model called GPT Image 2.5 Flare.
| Step | Before | Now |
|---|---|---|
| Ask for a picture | Open a separate image tool and repeat the idea | Ask in Copilot chat, in plain words |
| Edit it | Export, re-upload, and explain it again | Ask for the change in the same chat |
| Save it | Download it and move it by hand | Copilot saves it to your workspace |
What actually changed
The old flow had a clear gap. You had the idea in your code chat, but the picture lived somewhere else. GitHub says this update removes that context switch, so you can go from an idea to a usable image without breaking your flow.
It works in Copilot chat in VS Code, in the GitHub Copilot app, and in Copilot CLI. You can ask at the start of a chat, or right in the middle of a coding task. Your chosen reasoning model calls the image tool, the result shows inline, and you can keep working from there.
The catch worth knowing
This is not free in the way a chat message is free. GitHub says image generation is billed separately from your chat model, using image tokens that count toward your plan's AI credits. So a fun afternoon of picture experiments can eat into the same credits your coding work uses.
There are two controls worth noting. A visible "Made with AI" watermark is available, but it is off by default and you have to opt in. And owners of an organization or enterprise can turn image generation off through Copilot model policies.
Why I think this matters
For me, the win is small and practical. Mockups, docs, placeholder art for a landing page, and a quick edit to an existing image all fit this flow nicely. I still want a real designer for anything that has to carry a brand, and I still check every output before I ship it. But for the everyday pictures that sit inside a project, staying in one window is a real time saver.
Bottom line: if you already pay for Copilot Pro, Pro+, Business, or Enterprise, this is worth a try on your next small task. Just watch your AI credits, and turn that watermark on when the picture will be seen in public.
Would you trust a picture made inside your code editor for a real project, or would you still open a separate image tool first?
Cover image: AI-generated. Inside images: AI-generated.