Skip to main content
James373314849hdw
Participant
Oct 27, 2026 10:40 PM UTC
Question

Nano Banana models ALWAYS fail when using a reference image

For about a month, *none* of the Google Nano Banana models have worked when I supply a reference image in addition to the prompt.  I get a “Something went wrong. Please try again later” message EVERY time.

Other models which *do* allow a reference model work correctly.

I have tried clearing my browser cache, turning off my VPN, and disabling my browser extensions.  None of these things have any effect.

Please advise.

Edit: the Nano Banana models work fine *without* a reference image. They only fail when a reference image is used.

    4 replies

    N8.
    Community Expert
    N8.Community Expert
    Oct 28, 2026 6:45 AM UTC

    Sorry to hear you're running into that "Something went wrong" error ​@James373314849hdw . Could you provide more information; so, the team can proceed to troubleshoot? Without the requested information it is difficult to assist you and resolve the issue. By providing the requested info we can better assist you. 

    Something went wrong troubleshooting

    A Full Screenshot or short video:  Please provide a screenshot of your entire browser or application window, including your avatar in the upper right corner. Many errors have multiple causes; seeing the full context is much more helpful than a zoomed-in image of the error message alone.
    >>Make sure to remove person details: email address, name... before posting<<

    • System Details: Specify your Operating System (OS) and Browser (including version numbers).
    • Workflow Context:
      • Which Firefly function were you using? (e.g., Text to Image, Generative Fill, Text to Video, etc.)
      • Which model was selected? (e.g., Firefly Image 3, Image 4 Ultra, Imagen 3).
      • The full prompt you were using.
      • Copies of any reference images, keyframes, or original images used.

     For more details and additional troubleshooting steps, you also check this article as well!

    Resources & Suggestions on How to Prompt
    Generate images from text descriptions
    Writing Effective Prompts
    Adobe Firefly Resources

    Common problems

    • Trademarked / Copyright
    • Political figures / Celebrities 
    • Long complicated prompts (often from ChatGPT)
      (short easy to read prompts tend to work very well)
    • busy dark background images for videos. the images need to be easy to distinguish who the main characters are and what the background is. 

    Anything to do with Children

    • if it has children or child like images (an absolute no/ firefly will not generate anything that even resembles a child/ even childlike)

    or

    • - the prompt has words associated with children example: boy, girl, kid, kids, child, children and so on...
      Many Generation Models will not generate anything to do with children.
    • Nudity and adult content are strictly blocked.
    • For video models, avoid dark or busy reference images. The AI needs to easily distinguish the main character from the background.

    Cheers

    Nate

     

    Translate
    James373314849hdwAuthor
    Participant
    Nov 2, 2026 4:53 PM UTC

     I’m using the “Generate Image” function in Firefly. Mac OS 26.7.2. The browser is a “clean” copy of the latest Safari with NO user-installed extensions. I’m not using a VPN or other proxy.

    As I said above: for the last six weeks or so (I’m not sure exactly when the problem started), ALL THREE of the “Nano Banana” models return a “Something went wrong Please try again later” message when supplied with a reference image. (All the other models that can use a reference image seem to work correctly.)

    The reference image I’m using is a 3D rendering of a mannequin I made myself.  Be advised that Nano Banana fails for me on ALL supplied reference images.

    A lower-rez copy of the reference image and a screen shot of the browser are attached below. Please advise.

     

     

    Translate
      caroline_edits
      Community Manager
      caroline_editsCommunity Manager
      Community Manager
      Nov 2, 2026 7:57 PM UTC

      What happens if you specify what you want Firefly to do with the reference image? If you add onto your prompt that you want the illustration to match the pose shown in the reference image, does that work?


      Caroline

      Translate
      Terms & ConditionsAccessibility statement
      Using the community Terms of use Privacy Cookie preferences Do not sell or share my personal information ad choicesAdChoices