Google appears to be preparing a smarter, faster way to refine AI art inside Gemini. Evidence in a recent Google app build points to built-in markup controls for images generated in Gemini—specifically those created with the Nano Banana pipeline—so you can highlight exactly where you want changes without downloading, annotating elsewhere, and re-uploading.
What’s changing with Gemini’s new inline image markup
Today, Gemini’s markup tools show up when you attach an image for analysis or editing, a capability Google began rolling out late last year. What’s missing is parity for images Gemini creates itself. Strings surfaced in the Google app v17.8.59 suggest that gap is about to close: a pencil icon will appear on Gemini-generated images, opening a markup screen so you can highlight a region for targeted edits.
The flow looks straightforward. Generate an image, tap the pencil, swipe or outline the area to change, hit Done, and Gemini automatically adds that marked-up frame to your input box. From there, you enter a short instruction—“change the sky to sunset” or “remove the reflection on the glass”—and the model applies the edit using the mask you provided. It’s the same idea as inpainting, but natively integrated into the chat.
Why inline, region-based markup for Gemini matters
Reducing friction in creative loops is a big deal. Right now, many users bounce between Gemini and external editors just to indicate “fix this part,” which interrupts flow and increases the chance of miscommunication. With inline markup, you stay in one place, reduce steps, and give the model unambiguous visual context—often yielding better results with fewer prompts.
Region-based edits also help rein in overcorrections. Generative systems can be eager; a simple instruction like “brighten the subject” might alter the whole frame. A mask tells the model where to focus and where to leave things alone, improving precision. Competing tools have shown how powerful this can be: Adobe’s Generative Fill and Canva’s Magic Edit rely on masks to localize changes, and Adobe has reported that Firefly has powered over 14B image generations, underscoring mainstream demand for quick, targeted edits.
How this feature fits into Google’s broader Gemini stack
Gemini’s visual capabilities lean on Google’s imaging research, including the Imagen family for text-to-image synthesis. The “Nano” label typically denotes on-device or lightweight experiences in Google’s portfolio, and Nano Banana appears to be an internal codename tied to Gemini’s image generation workflow. Embedding markup at the UI level inside the Google app means edits can be orchestrated seamlessly, whether inference runs on-device for speed or in the cloud for quality.
This move also aligns with Google’s broader multimodal pitch: keep users inside a single assistant that can see, select, and change elements across text and images in one conversation. It mirrors the direction rivals are taking—OpenAI’s DALL·E supports masked edits, and platforms like Midjourney introduced region-focused variations—while leveraging Gemini’s chat context to stack iterative instructions without starting over.
Availability timeline and key things to watch next
The markup tools for Gemini-generated images are not yet live, and Google hasn’t announced timing. Given how features often land via server-side switches, expect a gradual rollout once testing is complete. When it arrives, look for a pencil icon on newly generated images in Gemini on mobile and potentially on the web.
Beyond first launch, watch for quality-of-life additions:
- Multi-region selections
- Feathering and edge controls
- A history panel for reversible steps
- Export options that preserve masks for later tweaks
Also expect Google’s safety systems to apply guardrails for sensitive content, consistent with the company’s responsible AI policies.
If executed well, this seemingly small UI tweak could have an outsized impact. Fewer context switches, clearer intent, and quicker iterations often translate into better output—especially for creators and casual users who prefer pointing to a problem over crafting a perfect prompt. Gemini’s upcoming markup-on-generation looks like a practical upgrade that meets people where they already are: inside the chat, mid-idea, ready to fix just that one thing.
