I’m getting an “An internal error has occurred” message when using Gemini Experimental 1114/1121 in Google AI Studio with prompts containing only images (one or more, no text).
Steps: Add image(s) to a prompt in Google AI Studio (no text), run, and get the error. Expected it to process the images or ask for text.
Workaround: Adding text to the prompt makes it work.
Anyone else seeing this? Is it a known issue?
Thanks!
notlin4
I tried to replicate it. Gave experimental 1121 just an image with zero text. It responded with
The image depicts a directed graph. It has four nodes labeled 0, 1, 2, and 3. Node 0 is a square while the other nodes are triangles. The edges are as follows:
0 -> 1 (dashed line)
0 -> 2 (curved line)
0 -> 3 (curved line)
1 -> 2
1 -> 3
It got the direction for two arrows wrong (it’s 2 → 0 and 3 → 0, not the other way around, and the arrows are crisp and clear), but it mostly worked.
Just in case diagrams are special and easy, I also tried an interior scene. No text prompt, just the image. The model nailed the description.
You probably ran into an intermittent internal error. Sometimes (recently) the internal error have been persistently intermittent and you need to retry a lot of times. Or, your image has something special.
I’m using Sample Media from the plus icon options.