Gemini felt dumb on September 19, 2025.

I tried to make a Polaroid‑style selfie with a celebrity on Gemini, but it kept refusing every single request because it’s a real person. Even my own friend’s photo got blocked. It’s maddening—totally dead ends on every try, no workaround, no clue why it’s so strict. I feel stuck, frustrated, and like I’m wasting time.

I played around with Gemini’s new System Instructions window, hoping it’d stop the endless adjective bloat and keep my story tight. It does help a bit – I no longer have to burn tokens or juggle 5k tokens of setup. But the game‑y “purple prose” still creeps back, and mimicking Claude’s tone feels more “just Gemini” than a true copy. I’m left frustrated that context drift isn’t really gone, and the model still drifts into bloated, generic sentences. I’m stuck waiting for a real fix.

I was trying to make a tall 9:16 image with Gemini, but every time the tool dropped me a square 1:1 picture. I kept telling it "ar: 9:16, tall image," yet the result stayed the same. It felt like a glitch that just stuck on defaults, and I was left frustrated and stuck with a useless output.

I finally enabled Gemini’s context memory feature after a week of waiting, only to watch it disappear the very next day. I spent hours hoping it was a temporary glitch, wondering if everyone else was seeing the same sudden drop. Now, a whole week later, I can’t get it back. It’s maddening to live with a promised feature that just vanishes, leaving me uncertain if it’s a bug or a larger system issue.

I’ve been playing around with Nano Banana and I’m actually happy with the photos it creates when I upload a scene, so I’m not totally rejecting it. But I’ve hit a wall with several requests that just don’t bite: when I ask to upscale to 4000x3000, the download is still stuck at 1024x768; when I try to get a full body, the head and feet get cut off; I can’t get a true widescreen or change depth of field at all; and focus requests are ignored. It’s frustrating that my instructions are being ignored, especially after getting promising results on other prompts, so I’m hoping there’s a trick or a setting I’m missing to make it behave as I expect.

When I first hopped into a week-long JSON-backed workflow with the AI, I felt lost and frustrated. I spent hours drafting a massive opening prompt, custom instructions, and a “state” file of my project. The first several rounds were clunky and misaligned. Slowly, as I tweaked the strict rules and added a “tunnel vision” mitigation rule, the responses became reliable. By the end, the tool’s output was crisp, consistent, and aligned with my vision. I could actually lean on it as a partner instead of a fickle chatbot. The shift from chaos to order was night and day – it saved me time, eliminated formatting errors, and gave me a powerful ally I could trust.

I’ve been hunting for a slick creative generator and hit Nano Banana everywhere, even on dating apps, but it’s a pain. It keeps “refusing” action‑movie prompts unless I sanitize the wording, while Seedream lands a first shot more often. Seedream’s shots come sooner, but Nano’s subject consistency is tighter. It feels like the safety net is rough‑cutting the fun—cheap for mainstream, but I’m left stuck rewiring prompts. Any reals on refusal rates, and where each really outshines the other?

I’m stuck trying to use Nano Banana inside Gemini; every time I click to edit an image, the prompt just pops up “Here is your image” but nothing appears. I feel like I’m missing a step or the tool’s broken—no output, no help. I’ve heard it works for others, but no luck for me; it’s frustrating and leaves me unsure how to begin or what to do next.