fathom hornet
Just like all other users who reported this bug here, I am encountering this problem since past few days. When I select gemini thinking model in gemini website and then select "images" from tool slider to generate image, gemini produces an image using nano banana 2, but when I select the option "Redo with pro" without disturbing anything, it generates text instead of image. This happens with so many prompts. When I asked gemini why is it not generating image when image option has been selected from the tool, it shows in the text result.
Steps to reproduce
- Open gemini on website
- Select gemini thinking model
- Select "create image" from the tool slider at the bottom left
- Write the prompt, then click the enter button:
Scan of a clean well preserved, very beautiful Persian miniature painting from Shahnameh of Shah Tahmasp, in strictly typical style of artist Sultan Muhammad, depicting a persian haunted house in a haunted persian garden with haunted trees and flowers at night. Colorful. The painting occupies the entire area of the miniature. For style of painting, use grounding with google search.
Aspect ratio: 9:16
(side note: please do not produce text response, this is an image generation task)
- When the image has been generated, select the three dots at left bottom and select "Redo with pro"
- Gemini won't use nano banana pro to generate image, instead will generate text result
Expected behavior: Gemini should use nano banana pro to generate image instead of wiring text.
grave merlin
main ore