Photoshop adds Nano Banana and Kontext Pro to Generative Fill

Adobe is giving Photoshop users more choice in their AI-powered workflows, today announcing that the Generative Fill tool will now support models from outside its own ecosystem. In an update to the Photoshop (beta) app, the company has integrated Google’s Gemini 2.5 Flash Image (aka Nano Banana) and Black Forest Labs’ FLUX.1 Kontext [pro], making them available alongside its own Firefly model.

This multi-model functionality was previously introduced in Adobe Express but is now making its debut in the company's flagship imaging application.

After 'selection' you can choose which AI model to alter the image. Image: Adobe.

The move acknowledges that different AI models excel at different tasks. Adobe suggests that users might turn to Google’s Gemini model for creating stylised graphic details, while Black Forest Labs’ FLUX.1 is positioned as a tool for generating elements that need to appear contextually accurate and in proper perspective within a scene.

Adobe’s Firefly remains the option for users seeking 'commercially safe, production-ready results with lifelike quality, giving creative professionals greater control over structure, style, camera angles, and zooms while enabling rapid iteration in up to 2K resolution for precision cropping, recomposition, and large-format printing'.

For creatives, this means the ability to select an AI model based on the specific needs of a project without having to leave the Photoshop environment.

A user could, for example, generate a background element with one model and then switch to another to add a stylistic flourish, before refining the entire composition with Photoshop’s familiar editing tools.

Nano Banana is currently one of the best AI models that can change peoples outfits and backgrounds – while adapting the lighting to suit the image. This image was altered directly in Nano Banana in Gemini. Image: Tim Levy
Nano Banana is currently one of the best AI models that can change peoples outfits and backgrounds
– while also adapting the lighting to suit the image (notice additional warm backlighting).
This 'straight from camera' original image (top) was altered directly in Gemini's Nano Banana.  Image: Tim Levy
You can check your AI credits left through your Adobe Account details.
You can check your AI credits left through your Adobe Account details.

Using Generative Fill, like the original Firefly, the new partner AI models will consume 'generative credits' that are allocated to users each month. Though unlike Firefly, users will only see 'one option / version' of the altered image instead of three.

The new Generative Fill options are live now for anyone with access to the Photoshop beta application with the Creative Cloud suite.

You can find out more about Adobe Create Cloud products on the Adobe Australia website.

How to use it within Photoshop (Beta): 

  1. Select the entire image using any selection tool, like the Rectangular Marquee Tool.

    Partner models such as Gemini 2.5 or Flux Kontext Pro use instructional editing prompts and work best with the entire image selected.

    This is different from how Firefly models work. With Firefly models It is best to select only the part of the image you want to add to or change.

  2. Next, Select Generative Fill from the Contextual task bar, select the Choose a model icon, then select Gemini 2.5 (Nano Banana) or Flux Kontext Pro from the model picker.

    Models have different strengths and generative credit costs. New partner models generate only one image. Firefly models generate multiple image variations and are commercially safe.

  3. Next, using the prompt field of the Contextual task bar, describe what you want to add or change and what you want to keep the same in your image.

    Use precise language with detailed descriptions for the best results.

    For example, you might prompt: Replace the woman's sweater with an oversized, fluffy blue faux fur jacket, keep her face the same. Add a pink neon sign behind her in cursive that reads "disco"

    For specific prompt guidance and best practices, visit the website for the model you select.

  4. Select Generate, review the results, and if needed, modify the prompt and select Generate again.

    Make big changes over several steps. Iterate your prompt language to improve results. If an image loses key details, start over, and describe the elements you want to preserve in new prompts.

Capture has already test driven Nano Banana and found it to be quite eirilly realistic. Just throw a bunch of elements into the chat and AI cobbles them all together. Of course sometimes it's mindblowingly realistic, other times it leaves a bit to be desired.
Being able to access the AI models within Adobe Photogshop helps workflow.  Image: Adobe.