thicc_girls_are_teh_best
Member
Okay, so the next part should be up on Saturday. It'll cover the 1993 period, but I'm not sure if I'll have the 1994 part up for Sunday. That's kind of also because, in addition to what I'll explain below, the year 1994 has a lot of things happening and I will probably even need to split it into smaller sub-parts (i.e one for the 1994 June Tokyo Toy Show).
I've been trying to get renders for some of the systems and it was a challenge to do that for the 1993 part. In part because I was being cheap and kept jumping between Gemini, ChatGPT, Copilot, Grok and some random AI image modifiers to try generating the look needed for some of the new devices. But due to all the limits I coughed up for a ChatGPT trail.
And well, out of those AI, ChatGPT is easily the best at following prompt constraints & directions accurately for image generation from what I've noticed, which was kind of surprising for me. Also the best at retaining original image quality (no artifact or blurring issues, etc.). Copilot is laughably bad no matter how much you simplify constraints, Grok never actually generated anything without hitting me with a message to subscribe, and Gemini was only "decent" at best. It would regularly degrade images over time when doing multiple modifications, and had issues with constraints that were directional in specific ways.
So uh yeah, brief update on that. For Project Tetsujin E (the NEC console), I actually have a rough sketch drawn up and hope to generate a proper real-life render from the sketch. You can see it below (top view), and I'll keep refining it and getting other angles drawn before trying to render it. But it won't actually show up until the timeline moves on to 1994
I've been trying to get renders for some of the systems and it was a challenge to do that for the 1993 part. In part because I was being cheap and kept jumping between Gemini, ChatGPT, Copilot, Grok and some random AI image modifiers to try generating the look needed for some of the new devices. But due to all the limits I coughed up for a ChatGPT trail.
And well, out of those AI, ChatGPT is easily the best at following prompt constraints & directions accurately for image generation from what I've noticed, which was kind of surprising for me. Also the best at retaining original image quality (no artifact or blurring issues, etc.). Copilot is laughably bad no matter how much you simplify constraints, Grok never actually generated anything without hitting me with a message to subscribe, and Gemini was only "decent" at best. It would regularly degrade images over time when doing multiple modifications, and had issues with constraints that were directional in specific ways.
So uh yeah, brief update on that. For Project Tetsujin E (the NEC console), I actually have a rough sketch drawn up and hope to generate a proper real-life render from the sketch. You can see it below (top view), and I'll keep refining it and getting other angles drawn before trying to render it. But it won't actually show up until the timeline moves on to 1994