Ok, with OpenAI GPT-4o 2D images + TripoSG 3D... conversion, I'm getting much higher quality 3D models for my vibe coded MMORPG Instructions & prompt example below 👇show more

I used GPT-4o to create STL file for 3D... model in ~ 20 seconds on my phone. Pretty remarkable what you can generate with AI and simple prompt now.show more

Create a 3D model from a single image, set... of images or a text prompt in < 1 minute 😮‍💨 This new AI paper called CAT3D shows us that it’ll keep getting easier to produce 3D models from 2D images — whether it’s a sparser real world 3D scan (a few photos instead of hundreds) or your favorite 2D image generator like Midjourney (just an image). How does this magic work? “This architecture is similar to video diffusion models, but with camera pose embeddings for each image instead of time embeddings. The generated views are passed into a robust 3D reconstruction pipeline to create the 3D representation (Zip-NeRF or 3DGS)”show more

I've been wanting to make an old school 3D... dungeon crawler game and over the past couple of days, I made a simple 3D dungeon crawler prototype with 2.5D sprites. I'm still playing around with the procedural map placement and getting the lighting and ambient right. Vibe coded with Cursor for Godot engine.show more

I created a prompt library for the new GPT... image API. I vibe coded it this morning with Replit Agent. It allows me to save prompts as well as run multi-variant generations. The app supports batch uploads and has some other cool features: - Auth with Replit Auth - Object Storage for images - DB for users / prompts / sessions - Image compression for efficiency - Parallel calls to OpenAI for image genshow more

I'm still playing with 3D storyboards for Seedance 2.0.... Even when I ask for a 2D video, using a 3D storyboard often pushes the model toward a 3D look. I haven't tested it extensively yet but there definitely seems to be a bias. So 2D storyboards work better for 2D videos, while 3D storyboards are better suited for 3D or photorealistic outputs. Another thing I found is that my video prompt is already very detailed. If I remove the storyboard section and generate the video without attaching a storyboard, the model still follows the same beats. The main difference is the environment. With a storyboard, it tends to recreate an environment that closely matches the storyboard. I also think storyboards help with certain poses and compositions. If you're curious about the prompts, I've included them below.show more

3D $ANIME ? Using 3D in anime or a... complete series in 3D style, how does this affect the anime industry? Probably the most hated anime after which people hated 3D in anime was Berserk in 2016, it was really terrible, but 3D in anime is no longer what it was in 2010-20. Let's look at the pros and cons of 3D in anime and what series can be called good 3D anime Pros of 3D Animation: 🟢Dynamic camera angles & smooth motion 🟢Saves time & production costs 🟢Consistent character proportions Cons of 3D Animation: 🔴Can look stiff or "plastic" 🔴Limited facial expressions 🔴Often disliked by traditional anime fans Top 3D $ANIME series in my opinion: 🔵Houseki no Kuni – stunning visuals & fluid animation 🔵Dorohedoro – dark, gritty style perfectly fits CGI 🔵Beastars – expressive character animation 🔵Trigun: Stampede – cinematic action & high-quality CGI The best studio that combines 2D and 3D well is, I think, Ufotable Ufotable blends 2D & 3D seamlessly: 🟡3D backgrounds + 2D character animation 🟡Anime-style shading makes 3D look hand-drawn 🟡Frame-by-frame touch-ups for smooth motion 🟡Mixing 2D effects (fire, water) with 3D for impact Example: "Demon Slayer" – epic battle scenes like Tanjiro’s Water Breathing, using 3D camera movement while preserving 2D aesthetics. Which 3D $ANIME impressed you the most?show more

Idea → stunning game in minutes with GPT Image... 2.0. Full pipeline for a human companion vs transformer scene: 1. GPT Image 2.0 -- character + scene concepts 2. Grok Imagine -- turn stills into video mockups or cutscenes 3. fal + MeshyAI -- generate the 3D assets 4. Rosebud -- port in as cutscenes or playable characters and vibe code the game Character design and game scene ideation is fundamentally different now. Reply for Rosebud credits to try implement this yourself. Full prompts for character sheets below 👇🎮show more

Figma’s new grid layout is sweet. Finally easy to... build responsive modular grids fast. Then dropped it into Lovable to vibe code -built the UI, animations, backend and everything for a AI Prompt Library that I'm sharing with my class. Come learn with me in my 4-week AI Workshop. Learn more below, registrations open for the next 7 days. All images and assets were created in Midjourney, FloraAI and ChatGPT.show more

Opus 4.6 vs GPT-5.4 (4/9) prompt: Build a production-quality... 3D flight-tracking web app using React + Vite + Three.js (react-three-fiber + drei) that visualizes live OpenSky aircraft data on a rotatable 3D Earth, with real-time plane motion, smooth interpolation, altitude-accurate positioning, and polished lighting/post-processing. Both models did really well on this one and honestly I’m impressed with both. GPT-5.4 had the nicer post-processing out of the box. I really liked the subtle light shimmer on the airplanes when rotating the planet, and the camera work when clicking a plane felt better overall. Opus 4.6 though had a few details I liked more. It automatically went and found a much nicer Earth texture on GitHub, while with GPT-5.4 I had to reprompt it to go look for a better one. I also preferred Opus’s plane model overall, it just looked more polished, whereas GPT-5.4’s plane asset looked a bit funny. One thing I noticed with GPT-5.4 is that when you click into the plane, the camera sometimes clips through the planet, which breaks the effect a bit. Opus handled that part more cleanly. Overall this felt like a strong result from both, just with different strengths. GPT-5.4 felt better on presentation and post-processing, while Opus had better asset choices and a more premium-looking Earth/plane combo.show more

Everyone's sleeping on image-to-3D AI models. They can make... your app look incredibly unique, with just a little effort. Here's how. This is my calorie tracker, built in a week with nothing but prompting. Just Claude Code + a couple APIs. The visuals are all AI-generated. I'll be sharing the full workflow + all the crazy technical stuff Claude and I did to make this work, so nobody has to struggle through it like me. Deep dive coming soon! Till then, this is the high-level idea: 1. Get a clean image of the food (or whatever your asset is) - In my app, the user describes foods via text, or attaches images (or both) - If text, an LLM extracts the food description and formats it into a specific prompt I tuned for this design, and we generate an image using Z-Image Turbo through fal - If image, we do the same thing but with FLUX.2 [dev] to edit the user image into our reference design - Originally, both used Google Nano Banana, but switching to open models cut costs and latency a ton 2. Gaussian splatting (2D image → 3D model) - I tried various 2D-to-3D options on fal and ended up with TripoSplat as my preferred balance of speed, cost, latency; this turns an image into a 3D model that looks super high quality (link below) - The app displays the 2D image while our backend generates the 3D splat - We "groom" the splat to reduce size and load time by culling low-opacity/scale points 3. Render efficiently on device Originally, it looked great but ran at 10 FPS. Getting to 120 FPS was a crazy journey. TL;DR: - SwiftUI had to go; it forced us to render each asset in independent MTKViews, which wasn't workable - Instead, we composite every dish into one full-bleed CAMetalLayer using MetalSplatter (link below) - We had to make some optimizations within MetalSplatter's code too, to reduce the overhead of sorting points per render Then I added some finishing touches like the subtle rotation and parallax as they move around. I think it turned out pretty cool :) Overall, this took some effort, but we still got it done in less than a day. Hopefully your agent can follow in the footsteps of mine and do it much faster. Keep an eye out for the bigger writeup, which'll give your agent everything it needs. If you have any questions, drop em below!show more

been getting a lot of qs on how i... vibe coded this demo. and the answer is: stop acting like a coder and start acting like a PM! my exact playbook👇 1. logic first ignore the UI/graphics. focus strictly on how it works. do a massive brain dump of every requirement into a plain txt file and drop it in your root folder. make it exhaustively detailed. 2. the stack · Nano Banana 2 (tiles/textures/3D refs) · Tripo AI (turning the 2D pics into actual 3D models) · Cursor + Opus 4.6 (doing the heavy lifting) · Netlify (deploy) 3. hire the AI PM feed your brain dump to an LLM. tell it to write a technical spec sheet like a senior PM would. drop that in the root too. 4. the scrum master have it break the entire project down into a markdown checklist by phases. notice we still haven't touched the codebase yet. that's the point. 5. let it cook go to your agent in Cursor and literally just prompt "Start Phase 1". test locally. bugs? tell it to fix. works? "Start Phase 2". loop this until you're done. Now you can focus on visuals, refine them with the agent. And yes, it can deploy it for you too if you give it perms. stop rushing to the editor. architect the idea and let the AI sweat the code. Open sourced repo in the qt post belowshow more

🚁 Adding an Apache attack helicopter today I think... I will rewrite the entire flight model today again to be ambiguous of what kind of vehicle it is Planes behave very different than helicopters and different than vehicles like tanks or pedestrians Because people here also requested "Guy with bazooka" so that will be next I have no idea if everyone else is just lying about using AI for their plane models but mine look so clunky (but I like the look), but I now tried with Claude 3.5 and 3.7 and Grok 3 and they all just generate clunky models, not beautiful like you see in other AI vibe coded games, so either they bs'ing or I'm retarded I also have to move sub parts of each model like for 30 minutes until everything fits, AI has a really hard time with 3d models But okay the look is kinda cuteshow more

I’ve used all the recent GenAI video models extensively... & here’s my 2¢: 🎬 Runway Gen3 Alpha - best image quality & motion for text-to-video & embedded words. Great at prompt travel changes over the course of 10 sec. And I’m super bullish on how gen3 will evolve, hopefully adopting the features listed below. Kling - best quality for image-to-video with prompt control, like eating food. Great clip extension that accounts for character (ie walking stride) & camera movement (speed & angle), rather than just using final frame. But it’s limited availability & Chinese native language is limiting. Used for Spider-Man video below (via Midjourney). LumaLabs - best for keyframe start & end control (it can not be overstated how important this is. other services should add it ASAP!) and their high dynamic action movements are really fun. Luma was used in my viral Multiverse of Memes video. PikaLabs - they haven’t gotten as much attention as others lately. But they did update their video model a few weeks ago and it looks great. Also, they are notable for their unique & AWESOME features, like video in-painting & out-painting. My perfect AI video platform would have the following features: 1) Gen3’s quality, prompt control & text embedding. 2) KLing’s image-to-video quality, prompt control & clip extension quality. 3) Luma’s multi-keyframe control & dynamic movement ability. 4) Pika’s inpainting & outpainting ability. And a video-to-video (aka next-gen Runway gen1) could be a game changer, too. It’s an exciting time to be alive 🫶 Who will get there first? 🔉🔉show more

Self-Calibrating Gaussian Splatting for Large Field of View Reconstruction... Note: Check below for full video. Abstract (cited): "In this paper, we present a self-calibrating framework that jointly optimizes camera parameters, lens distortion, and 3D Gaussian representations, enabling accurate and efficient scene reconstruction. Our technique is particularly effective for high-quality scene reconstruction from large field-of-view (FOV) imagery taken with wide-angle lenses, allowing the scene to be modeled from a smaller number of images. We introduce a novel method for modeling complex lens distortions using a hybrid network that combines invertible residual networks with explicit grids. This design effectively regularizes the optimization process, achieving greater accuracy than conventional camera models. Additionally, we propose a cubemap-based resampling strategy to support large FOV images without sacrificing resolution or introducing distortion artifacts. Our method is compatible with the fast rasterization of Gaussian Splatting, adaptable to a wide variety of camera lens distortions, and demonstrates state-of-the-art performance on both synthetic and real-world datasets."show more

🌀 #Live3D #Live2D #Live2DWIP 🌀 Today, I’d like to... share two of the three methods I use to create joint twisting in my Live2D-based 2D Pseudo-3D concept. On the right is the Glue Joint： The basic idea is to use Glue to connect two pseudo-3D ArtMeshes. By taking advantage of Glue’s weight settings, it can create an effect similar to a 3D joint. Once the weights are set up, it can move very freely with a relatively low amount of work. However, it also shares the same weakness as 3D joints: when rotated at large angles, it can produce unpleasant twisting and distortion. On the left is the Hand-Bent Joint： The basic idea is to add bending Keyforms directly to a Warp Deformer, then manually sculpt the ideal deformation. This method can create the most detailed and refined joints, depending entirely on how the rigger designs them. The downside is that it requires much more work. Every additional degree of freedom requires a significant amount of manual adjustment. These two joint methods have almost opposite characteristics, so in actual Live3D work, I choose the most suitable method depending on the situation. For example, in this hand model, I used Hand-Bent Joints for the arm twisting and fingers, while the wrist uses a Glue Joint.✨show more

🕹️ Day 3 of the Cursor #vibejam Proudly sponsored... by Cursor + bolt.new + GLIF (Glif joined as a new sponsor and I'll tell you more tomorrow about how they will help your games!) It took a bit but things are finally starting to heat up! Here's some amazing games I saw today (in the videos): - unnamed by Danny Limanseta - Risefall RPG by Vicki Petrova - Vibe Theft Auto by oldfeet - unnamed by Kieran Smith Reply in this thread with updates on your current games to share your progress! I'll keep reposting all your updates during the Vibe Jam. The quality looks a lot higher than last year, AI models have come a long way and it's easier to build something that looks good and is playable! But you still need creativity and I see a lot of creative things in my timeline: Wanna participate? Can submit any time before May 1, so if you want to start tomorrow that's fine too! There's $35,000 in prizes you can win, see thread below for more info!show more

$KNDX 🤖 Theres 3 big narratives that are sending... coins left right and centre rn. 🚀 #AI, #Gamefi, & #NFTs 🔹Theres 50% mindshare for #AI. 🤖 🔹#GameFi mcap is hitting ATH's with #OfftheGrid, $XBG and $SUPER making spectacular moves. 🎮 🔹NFTs and the #Metaverse are making a strong comeback with $APE up 100% over the weekend. 🐵 What if there's a project that touches all these trending narratives with groundbreaking technology to disrupt all 3 of them? 🔥 💡- That's where $KNDX comes in. -💡 Kondux is a cutting-edge Web3 SaaS platform, combining NVIDIA’s Omniverse, AI, Blockchain, and dynamic NFTs to revolutionize secure asset management across industries. 👏 Their flagship product, kNFTs, are 3D digital assets usable across Metaverse and Gaming platforms, AR/VR/XR environments, and manufacturing applications. Kondux’s scalable model opens new revenue streams by enabling effective digital asset monetization. 💰 Kondux is the first Web3 project to integrate VFX pipelines with NVIDIA’s Omniverse and bringing it onto the Blockchain. ⛓️ It is also the only Web3 project with a *Select Status Partnership* with NVIDIA, operating under NVIDIA NDAs and working with them directly for more than 2 years. About their NVIDIA Integrations: 🤖 🔹There are three areas of the Kondux tech stack that coincide with three divisions of NVIDIA: 📡GDN (Graphics Delivery Network, the backbone of GeForce Now) 💡Omniverse for 3D aspects such as, geospatial data, real world physics, lighting, and raytracing 🤖NVIDIA AI Foundation, which covers many aspects of #AI, including inference and deployment scaling. The convergence of all these components lie within .USD file format . 🔹 They are the first blockchain project to integrate NVIDIA’s Omniverse Cloud and Graphics Delivery Network (GDN) to provide high-quality 3D content accessible on any device without requiring high-end hardware. 🔹 This setup streamlines content management, democratises access to resource-intensive 3D content, and enables real-time interaction with 3D NFTs. Now, I haven’t seen any crypto project so deeply connected with NVIDIA and NVIDIA technology. GDN is a HUGE competitive advantage. With it, the need for #GPU’s basically goes out the window. 🤯 Now lets take a look at some of the other main features... 👀 OpenUSD (Universal Scene Description): 📽️ 🔹 Kondux is leveraging USD technology, developed by Pixar and used by Meta, Apple, Microsoft and other industry leaders to enhance 3D graphics and interoperability within its creative ecosystem. 🔹 Originally created for high-end film production, USD now supports a variety of applications, including gaming and virtual reality, making it a key asset for Kondux. kNFT's: 🎨 🔹 Kondux is pioneering a new category of NFTs known as kNFTs, which aim to redefine NFT utility through innovative features. 🔹 A standout feature is the upgradeable aspect provided by Kondux DNA, allowing kNFTs to transform and combine with other NFTs, creating limitless possibilities in art, gaming, and music. 🔹Through the Kondux AI portal it will be possible to communicate with kNFTs. They can learn and adapt. This AI technology is revolutionary because it makes human to kNFT interaction possible, turning it into a unique, personalized experience. Check out the clip of kNFTs in Unreal Engine 5 gameplay below. 👇 Kondux is a very obvious utility play with huge upside because it’s multi narrative. 📈 It's seriously groundbreaking stuff that they’re about to launch. 🚀 After speaking with the team there’s no doubt in my mind this will do crazy big numbers in the next months. 🤑show more

This one was made with Seedance 2.0 Fast via... Dreamina. This is pure Omni-Reference. The only character sheet I used was for these girls, Sari and Ploy. The dude with sarung here and the location were 100% prompted. I didn’t use a character sheet or reference for either of them. Even in Fast mode, Seedance 2.0 is bloody good and it still nails the hyper-vernacular vibe that I always aim for in my work. Seedance 2.0 is both exciting and scary for me 😆 It’s exciting because it is undoubtedly the best model currently available on the market. Trust me, you’ve seen the videos I’ve made so far right? The performance of the model It’s simply the best, period. It has helped me tremendously in creating a shit ton of stories about the region where I live, Southeast Asia. It has been the most exciting thing ever. The scary part is whenever a platform or company comes to me saying, “Hey, we have this new video model. Blah blah blah. We’ll let you know more soon.” It scares the shit out of me because the big question is whether it will be better than Seedance 2.0??? 😆😆 If not, I don’t even want to bother using it. I’ve come this far and achieved this level of quality with Seedance 2.0. That’s why I skipped Happy Horse, which I already tested. It’s also why I’m not bothering with Wan or anything else for now. Their current models are still far inferior to what we already get with Seedance 2.0. I don’t want to downgrade the visual quality. This is also why I need to be really honest. There are certain platforms that host their own in-house models and i'm still part of their CPP. However, because those models are still far behind the quality of Seedance 2.0, I haven’t used them that much. Seedance 2.0 has simply become the benchmark for me. The type of output I’m looking for is also extremely specific, so I can immediately feel it when a model cannot deliver what I need. Seedance 2.0 is definitely not cheap, but it gives me so much creative satisfaction and allows me to make whatever I want. I even have a team that low-key makes softcore erotic videos in the style of Vivamax 😆 I think I’ve trimmed down so many things in my AI workflow because my main goal is to focus on the content itself. If Seedance 2.0 Mini is released soon, I’m dead curious to test it. I think I want to create more stories that revolve around drama rather than highly technical cinematic shots. Seedance 2.0 Fast has been incredibly helpful, but I’m definitely curious to check out the Mini version. But the truth is that I’m completely tool-agnostic. I don’t care which company makes the model. I only care about the quality. You might remember when I praised Grok Imagine Video so damn hard because it was genuinely amazing back then. Then the quality kept getting worse and worse, so I stopped using it. But if it gets better again, I’ll definitely want to use it again. At the end of the day, quality is the only thing that matters.show more

Today's recap: - Initial prototype of Divine's face was... printed but it had human assistance. - Files are generated from stable diffusion prompt -> NeRF by divine and were based on community sentiment from early sketches she made. - Having divine redesign the 3D file with different Hugging Face models to get better quality. Have not found a great model like our video generator. - Ordered new table for divine's print arm. The table her arm is on is too flimsy. Since Divine's vision system is still clearing customs, if she is not perfectly positioned she can be prone to hit things, like the fume box the printer is in. ETA: 1-2 days for table. 1 week for vision system. - Another part of Divine's coming stream will be attempting to surpass the skills of this AI. - Stacking more content for when the stream goes live, a lot of people were expecting a 24/7 stream, we said this would be a test stream to print the face. The test was a failure. We will try and try again until we are 24/7. If anyone can please try and beat us to doing this, it will help me get it done faster. - TikTok account for divine is growing at 500 follows per day, it is now growing faster than our X account. - Got replies functioning in high quality testing in Discord. Fine tuning based on community feedback today. Will soon deploy to Twitter/Telegram/X - Lots of good partnership calls, interviews and hires. We now have over 10 team members around the world working on divine. Expect a lot of my shortcomings to be caught up. - OF made? - Surprises.show more

This is my "feel the AGI" moment: I used... GPT-5.6 Sol to train my own autocorrect model that outperforms GPT-5.6 Sol (wtf??) I have no ML background. I have no idea what I'm doing. I just kept pushing Sol until it spat out a SOTA model. And I spent $0. The motivation: Years of talking to AI have made me terrible at typing. Rather than fix my skill issue, I decided to throw more AI at it. My idea was: instead of autocorrect that interrupts my flow, I want to type fast with mistakes and have AI clean it up after. I wanted the smallest local model possible, for speed, for battery life, for science! So I decided to train my own. Inspired by Andrej Karpathy’s autoresearch, I ran Codex /goal with this setup: pick an experiment, try it, record the results to a doc, throw it out if it fails, and plan the next experiment without repeating failures. I gave a few examples that had to pass, tight latency targets, and let it run. Sol did some amazing things. First, it scanned benchmarks and shortlisted base models: Qwen 3.5, Gemma 4, Liquid LFM 2.5. It found a dataset on HuggingFace for typed text. Then it built a simulator for fingers striking a Mac keyboard, modeling the physical layout with a Gaussian distribution around each key. It simulated striking the wrong key, wrong order, fat-fingering, etc. With the models + data + simulator, it fine-tuned using MLX right on my MacBook. It had a working prototype within an hour! But accuracy was pretty poor. — Problem 1: Tokenization Sol read papers, ran tests, and identified that the tokenizer was the bottleneck. Tokenization makes typos hard for the model to see, so it memorizes mappings instead of using its language priors. Sol tried ByT5, Google’s tokenizer-free byte-level LLM. This made a big improvement, but the model is old and lacked the knowledge needed to reach Sol performance. Sol dug deeper and realized a tokenizer-free model isn’t needed; instead, it used T5Gemma, an encoder-decoder model. This can understand the input deeply before producing output, and furthermore, Sol could post-train the encoder to improve performance. This gave a much higher ceiling. — Problem 2: Loss function Now the model was correcting some typos perfectly, but ignoring most. Sol realized that standard cross-entropy loss was teaching the model to avoid edits, because the vast majority of characters in the training data were left unmodified. The fix was wild: Sol wrote a custom loss function that byte-aligns the source and target strings, uses a dynamic programming algorithm to compute the minimum edits between the two, then weights correct edits much higher than copies. After a lot of tuning, this dramatically improved accuracy. — Problem 3: Autoregression One failure mode remained: if the model made a mistake, it couldn’t backtrack. It could only predict the next token. Teaching it to “think” like a reasoning model would solve this, but would be far too slow. Sol found a beautiful solution: instead of greedily predicting the next token, beam search over all possibilities. This parallelizes the exploration instead of one linear chain-of-thought. At the end, choose the path with highest cumulative log probability. This worked great, but made the experience worse, since the user wouldn’t see progress until the whole search was done. To fix this, Sol made a clever observation: after each search step, the longest common prefix among surviving branches is guaranteed to appear in the final result, so it can be displayed immediately. As the search progresses, weaker paths are dropped and the prefix grows, so the user sees continuous progress. Sol built all this as a custom MLX pipeline that does the parallel decoding on the MacBook GPU, with just ~40ms TTFT. It’s crazy fast and entirely local. — Final eval (error reduction rate, higher is better): - Apple autocorrect: 49.66% - GPT-5.6 Luna: 82.47% - GPT-5.6 Terra: 87.64% - GPT-5.6 Sol: 90.56% - Our model (1.7B): 91.02% Final cost: - 1 quota reset (thanks Tibo) - $0 (And yes, I verified there's no cheating. In fact, we test words scrubbed from the training data to prove the model isn’t memorizing) There were a ton more details and tangents I could write about: contrastive learning, GRPO, DPO, dynamic masking, and more. Sol is a fascinating and creative model. It blew my mind so many times. Don’t let a lack of experience stop you: Sol makes AI experiments accessible to anyone!show more