After the update, I see that AI Studio either refuses to generate images now saying that it’s a text-only model or generates a json with the image description and nothing else.
Before you could just generate and ask to modify images, is this feature broken now or has it been disabled for free users?
Hi @berryberry,
Welcome to the Google AI Forum!

To better understand the issue, can you please share the following:
- Model name?
- Configuration settings (thinking, grounding, etc)
Also please share prompts and responses here for better assistance to help us reproduce from our end.
Model: Gemini 3 Pro, selected the Image option on chat start
Thinking is set to High, grounding is Enabled (I left all settings on default)
Example prompt:
Generate an image of a banana wearing a costume.
Response:
{
"action": "imggen",
"arguments": {
"prompt": "A whimsical, photorealistic close-up of a banana dressed as a superhero. It is wearing a tiny red fluttering cape and a small blue eye mask. It stands triumphantly on a wooden table. Dramatic cinematic lighting, high detail, 8k resolution."
}
}
and no image.
Thank you for clarification. I was able to successfully reproduce the issue. I will escalate this to the concerned team.
Thanks for sharing the bug.
Have a nice day 
1 Like