Has anyone have been having issues with the Gemini live api multimodal input when it comes to streaming video live? The model seems to not know what’s in the image anymore for example I ask what color shirt I am wearing it says blue instead of grey. I also tested this with the boiler plate code google provides and same-thing.
1 Like
Today video mostly doesn’t work for me. The model will hallucinate if pressed but can’t see anything.
2 Likes
Yeaa it started late last night for me, I’m wondering if it’s because they’re updating something.
1 Like
We are looking for facts 
strong text
Continued to mess around with this and for some reason it detects images when using audio input and not text input…weird