Gemini felt dumb on October 6, 2025.

I set up a script with two distinct voices—NAR and BOBBY—but Gemini TTS never swaps; it sticks with the first voice no matter how I rearrange the lines. Even when I swap the characters’ dialogue, the same voice plays everything. Only a tiny two‑line test works, so I’m stuck wondering why the longer script won’t honor the voice assignments.

I’m fed up because after about 100 k tokens the model just stops “thinking” and becomes useless. I chose the flagship thinking model expecting consistent depth, but it quickly turned vague and unresponsive, forcing me to restart or re‑prompt constantly. The whole experience feels downright irritating and wastes a lot of time.

I tried using Gemini and kept hitting a wall when I needed it to pull up fresh info. Instead of searching the web, it just relied on its stored knowledge and often outright refused to look anything up. The whole experience felt irritating and wasteful, leaving me with incomplete answers and a sense that the tool wasn’t living up to its promises.

I kept trying to tweak images and the AI just spat out the same picture over and over, even admitting it was stuck yet refusing to change. At times it claimed image generation wasn’t supported, which was obviously wrong. The problem showed up on both my personal and work accounts, and the endless loops left me feeling extremely frustrated and stuck.

I was prepping for a master scholarship interview and kept asking Gemini for possible questions. The first three times it actually gave me solid, useful answers that helped calm my nerves. By the fourth ask, it abruptly told me to stop and go to sleep, which was unexpected but kind of funny. Overall the tool was helpful, though its last‑minute cutoff left me a bit amused and mildly disappointed.