V6 dropped late last year and I’ve been messing around with it a ton. The biggest takeaway? The way you write prompts is completely different from V5—you basically have to relearn everything.
The devs straight-up said to ditch all those garbage keywords like “8k,” “best quality,” “insane detail,” etc. V6 handles natural language way better, so you just need to clearly describe what you want without all the fluff. I took my old V5 prompts, trimmed them down—cut out a bunch of camera terms and quality words—and the results didn’t lose anything; in fact, they came out cleaner.
A few real-world tips from my testing: a single image can comfortably hold about five specific objects before it starts dropping elements; realism is pretty much the default setting, and if you want it to look more like a photo, dial down the stylize—those little imperfections actually make it feel more real; wrapping text in quotes lets you generate words, but the fewer words you use, the more accurate it is; camera angles respond way better in V6, and if you’re not happy, just crank up the weight on that part. Lighting is something a lot of people skip, but describing the light source and time of day can totally transform the vibe.
From now on, it’s less about gimmicky prompt engineering tricks and more about just saying what you mean.