Building a real-time SDF+AI creative tools @unbound_io with @andreintg. 🇺🇸🇦🇹🦘- @speedrun SR004 -If you feel any nihilism wrt gamedev: stop it. get to work!
San Francisco, CA
Joined January 2009
- Tweets10.5K
- Following547
- Followers2.4K
- Likes7K
How about giving the creators of content on digital stores a bit more surface to showcase who they are? For customers who care?
We totally lost this in Spotify, Netflix etc.
Free form perhaps? Show your process, you office, your team? Concept art?
In essence: the CD booklet
Florian Hoenig retweeted
Developers have always had the option of creating their own engine. Now it’s going to get a lot easier, especially for games with simpler needs. There are also permissively licensed alternatives like Godot. Everyone should do whatever is best for their particular game.
Cool emergent thing from "wobble-man" mode is that he's trying to hold on to trees if you hit it at the right angle.
Yes, this is real. Macy isn't over-promising here.
speedrun founders land Customers most startups wait years for. Here’s how our GTM team gives you that unfair market advantage.
Trust me, watching this 3 min video can get you over $10M+ in credits, but you gotta apply to @speedrun too: speedrun.a16z.com/apply.
Real-time augmented SDF sculpting for game assets.
It was a pleasure to work with the @tripoai team to get this done at launch.
You can try it in loop.unbound.io today!
EdenLive is now powering @unbound_io's AI-augmented sculpting workflow.
Bring real-time generative rendering directly into your 3D creation loop. Sculpt models, watch AI refine them instantly, and export game-ready assets with Tripo.
Generated with Tripo. Rendered in real-time with EdenLive and Unbound.
An initial wave of competent OG programmers are on it at this very moment. And it’s not even all of them yet.
Respect to you Gabriel, but you certainly
gonna get more than one ;-)
We have collectively put a lot of effort into having our culture develop acceptance for moments in our lives where your friend John is now Johanna.
Cool.
Why is it so hard to accept that your friend who was passionately writing machine code since they were 7 years old now vibe codes like a motherfucker?
🤷♂️
Another great example of what's coming to the modding scene.
Pantelis is an extremely competent engineer but probably would not have had the time to pull this off a year ago. Now it can be a casual hobby.
Bloodborne VR is out now :) The whole game has been redone to play and feel like a native vr game.
Get it here: github.com/bldbrne-vr/bldbrn…
Arrived at the Istria 🇭🇷 biome.
Took 15min at jet speed mostly held up by evaluating the ONNX terrain model.
Distant forests always look a bit like "Thanks for rubbing your LOD into my eyes." A bit of contour smudging seems to resolve this problem and things look a bit more like you actually wanna go explore.
(no AA present yet)
It's absolutely hilarious how often the best coding agent messes up cascaded shadow maps. Anthropomorphizing say: "I feel you little buddy".
Reality: there must be a lot of crap implementations in the training set and reasoning oversees it. nobody ever enjoyed implementing it.
Florian Hoenig retweeted
Experimenting with low-fi asset gen in Unbound Loop. Mostly focusing on semantic parts splitting done at generation time, because 3D gen segmentation sucks.
Would anyone be interested in giving this a try? 🤔
it’s already in your timeline, apologies. however we all need to speak up against this right now.
It’s time for a million dollar gaming console. Why spend money on game development when your players are all rich AF. 😂
🚀 Excited to introduce StreamRender-H3 — bringing real-time interactive worlds to life with diffusion models!
Built on Sol-Engine, we post-trained @MiniMax_AI H3 Ref2VA into a real-time streaming neural renderer.
🎮 Code drives the world. A game engine handles physics, object states, and player interactions, while H3 generates the visuals.
⚡ Real-time streaming: 2 denoising steps, 24 FPS, and ~273 ms per chunk on 8×GB200 GPUs.
🎨 One world, endless appearances: Change the reference image and prompt to render the same interactive world in entirely different styles.
🔥 Powered by streaming post-training, distillation, a fast causal H3 VAE, and full-stack inference acceleration.
Check out our demo and code! 👇
demo: nvlabs.github.io/Sana/Sol-En…
code: github.com/NVlabs/Sana/tree/…
Opus 5.5 "/goal" on a set of reference images: 1800s romantic landscape paintings.
In the morning it was repeatedly stating "cannot reach the goal". Fine :-) Results are still pretty cool. But it does not look like Alps.
I'm confident I can iterate towards that though.
Next up is to redo the procedural flora. It did a broad sweep on 50 something species from my design document. They are placeholder that need to be refined with reference image.
However the reference images need to be of uniform style.
Now also various ages of each type.
