@Lucas09aoht@LexnLin OpenAI optimized its models for efficiency that translates into real-world work.
Anthropic optimized for squeezing every last benchmark point out of a mountain of tokens.
Very different philosophies.
Unlimited money? Opus 5.5.
Living in the real world? OpenAI.
@thsottiaux@omarbnjl How long will it take for the resets to propagate? It's 2am here in southeast Asia and I was wondering if I should sleep tonight or not
Tibo, mate, wake up. A million devs are waiting and somehow the AI revolution is bottlenecked by your index finger. Whack the fucking button monsieur @thsottiaux
That's great! I'd love to bring some of my games to a Japanese audience too. I've visited Japan quite a few times and absolutely love the country. If there's ever an opportunity to collaborate or help each other bridge the JP/EN side of things, I'd definitely be up for it. Looking forward to seeing where you take the game!
I haven't tried Godot yet. I usually work with Phaser or Unity depending on the project. Astra + Codex + Image2.5 work extremely well together. Looking at your work, I can see a lot of similarities in our style, output and approach to design. Good luck with the game! I'm excited to play it once it's ready. I might need to brush up on my Japanese before then though ๐
This looks good. I'm not a 3D pro, my background is in software engineering, but this already looks pretty decent.
I'm sure there are plenty of opportunities to improve it, and efficiency is something we can optimise once the workflow itself is solid. The way I see it, you first have to prove the core flow works reliably, then you can start tackling the economics.
Astra + Jev + Blender + Image 2.5.
For simpler character models, like the ones I'm using in a game I'm building, this already works pretty well and can be quite economical. But right now, it only works reliably at a fairly simple level.
Feel free to DM me if you want to dig into it more, would be cool to compare notes.
Yeah, native Astra is good with simple shapes and excellent with architecture, but character modelling is still pretty rough.
I built a plugin to fix that part of the pipeline, but I need more usage to finish it properly. It still can't handle fur or hair, and this is just a single demonstration and it took a bit of trial and error to get here. Making it reliable across every kind of character will take time, and I want it to be able to one-shot image to 3d pipeline at Meshy level.
The economics are a completely separate problem I haven't touched yet.
But based on the results I'm seeing?
Meshy is doomed.