>>150823751
I don't want it to seem like I'm down on gemini; I think it's amazing. I did the figurine thing like everyone else, but when I got a handle on how to prompt for it, and how good it was at manipulating images, (like the big library of dall-e gens we all have at this point) suddenly all these new possibilities were open. I haven't had as much fun(or been as addicted) with a generative tool since the first month of dall-e.
the fact it IS so amazing and powerful is what frustrates me so much with stuff like that, because I know it CAN do it, it's just getting mixed up somehow. It has this one particular tendency which drives me crazy, I've seen it a lot. You'll ask it to do, say, two lines of subtitles, and it will nail it except mess up one word badly. You'll point it out and ask it to correct it, and a lot of the time it will start panicking and go into a complete tailspin, doing everything imaginable BUT fix the one word.
do it again but misspell it slightly differently(and it is always the same word it fucks up), randomly use an image from higher up in that chat, overlay other text on top of it, change the style completely, paste an old image over the top of the text, correct the text but fuck up something else in the image, (my favourite) post a compete non-sequitur new image that has nothing to do with anything, etc. it will do everything BUT fix one typo in one word.
its comprehension is fantastic, it always seems to understand what I want it to do, or what it is meant to be fixing, it just has this weird blind spot that irks the shit out of me.
but this is like 1-2% stuff i'm complaining about. it's just the juxtaposition of its amazing capabilities that would have seemed like sorcery even 5 years ago, and it still struggling to fix a typo, something computers have been able to for fucking 50 years
I endorse gemini
everyone should at least try it
if you gen in any other tool you can surely find use for it
