Gemini felt mid on November 9, 2025.

I keep trying to get Gemini to produce 9:16 or 16:9 images for Stories and YouTube, but it stubbornly returns only 1:1 squares. I experimented with every prompt tweak I could think of, and nothing changed. It feels like a hard‑coded Bard limitation, so I had to switch to Twills AI. It’s not free, but it finally lets me generate high‑resolution images in any aspect ratio I need.

I’ve been chatting with Gemini and noticed a clear drop-off after the first couple of replies. The opening answers feel sharp and insightful, but by the fourth or fifth turn the responses become vague, generic filler that sounds technically correct yet hollow. The repetitive praise (“you’re right,” “great insight”) feels forced and irritating, making me think the model is throttling its compute as the conversation drags on.

I tried to use Whisk to create a consistent character head, saved a shot I liked, then fed it back as a reference. Instead of recognizing its own work, the system started rejecting those images left and right. It’s become so over‑broad that even its own creations are flagged, effectively rendering the tool unusable. I had to roll back updates just to keep it functional.