Been experimenting with something a bit bigger this time 👨‍🍳 Instead of image → one 3D model, I’m trying image → entire 3D world. Break the image into objects, rebuild them, reconstruct the scene, then make it interactive. Would love to hear what you guys think about this direction for img2threejs.

This was ONE image. Now you can walk into it. ❄️ Not just the cabin. Not just the character. The whole scene. Hyper3D × img2threejs × Three.js × GPT-6 Luna The workflow starts with a picture of a world. img2threejs analyzes what’s inside: the house, trees, rocks, character, and their arrangement. Hyper3D provides the 3D references. GPT-6 works through the img2threejs pipeline to reconstruct the objects and bring them together in Three.js, rebuilding the scene with depth, lighting, and interactions. But recreating the picture is only the beginning. Walk around the cabin and leave footprints in the snow. Open the gate. Turn day into night. Place a campfire and watch the world take on a different atmosphere. You’re no longer limited to the moment or camera angle captured in the original image. This isn’t just making a picture move. It’s rebuilding what’s inside it into a world you can explore, edit, and keep building. One image sets the scene. You decide what happens next. Opensource: @DeemosTech @threejs @OpenAIDevs #img2threejs# #opensource# #imageto3D#

GPT-6 × img2threejs v2.2 GPT-6 launched while I was halfway through building img2threejs 2.2, so I ran an early comparison between Luna, Sol, and Astra. Same reference image. Same prompt. Same unfinished pipeline. Skill: “Use the img2-environment skill to recreate the room from this reference image: [link]” v2.2 is currently around 50% complete, but the differences in how each model interprets the same scene are already clear. Sol’s visual understanding stood out to me in this test. The goal for v2.2 is to push reference-to-scene fidelity as close to 99% as possible.

GPT-6 launched right in the middle of building img2threejs v2.2. As its creator, I had to test Luna, Sol, and Astra using the same image, the same prompt, and the same unfinished pipeline. img2threejs v2.2 is only 50% complete, but Sol’s visual understanding genuinely surprised me. GPT-6 launched right in the middle of building img2threejs v2.2. As its creator, I tested GPT-6 Luna, Sol, and Astra with the same reference image and one simple prompt: “Use the img2-environment skill to create a room that matches the following reference image: [link]” v2.2 is only 50% complete, but Sol’s visual understanding genuinely surprised me. Which result do you prefer: Luna, Sol, or Astra? Which result do you prefer?

Sharing a small game character experiment I just built with img2threejs. I used the recent Sora & Roxas × Fortnite release as a fun test: starting from reference images, generating editable Three.js characters, then putting both on a compatible rig system with 44 shared animations. They can even switch characters mid-animation without resetting the motion. The interesting part for game dev: the result is code-first, but I’ve also added GLB/FBX export, so the assets can be taken into Blender or Unity and continued in a traditional game pipeline. I’d love to hear feedback from game devs here, especially if anyone wants to test the exported assets in a real game workflow. Full showcase in the quoted post. 👇

A single image became a fully animated 3D character built as editable Three.js code. Sora and Roxas just arrived in Fortnite, so I rebuilt both with img2threejs. One compatible rig - Two characters - 44 shared animations. They can switch mid-animation without resetting the motion, then export to GLB for reuse in Blender. Image → code → reusable 3D. Riku or Kairi next?

I’m sharing a code-only Three.js showcase in which the 3D model, materials, lighting, and animation are all created and controlled directly through code. No GLB file is used or included in the final output. This keeps the scene fully editable and makes iteration easy for example, adding three or four new iPhone Duo color variations with a single prompt.

img2threejs turns @Apple 's iPhone Duo references into interactive 3D experiences with ThreeJS. A 2D image can show a product. Interactive 3D lets people rotate it, unfold it, customize it, and understand it before they buy. Live showcase: The img2threejs pipeline makes this practical: 1. Provide product images, proportions, and dimensions 2. Generate an initial 3D study with Hyper3D MCP 3. Use prompts and visual feedback to refine geometry, materials, and details 4. Add custom colors, screens, camera views, interactions, and animations 5. Review everything directly in the browser and continue refining Apple introduced two official iPhone Duo finishes. With img2threejs, I explored four more custom concepts: Sage, Coral, Sand and Lavender. The result feels less like a render, and more like a product people can understand, explore, and love. Full workflow breakdown on YouTube in about an hour. 🚀 #img2threejs# #GPT6# #iPhoneDuo# #Hyper3D#

A BIG update for img2threejs. It’s been almost 2 months since I launched img2threejs. After a lot of experiments, feedback, and seeing where people struggle, I realized there are some changes we simply have to make if we want this project to go much further. So here’s what’s changing: / 01 / A new ROADMAP: from one pipeline → a plugin-able ecosystem This is the biggest change. Reconstruction, environments, rigging, animation, VFX/SFX, exporters, providers… shouldn’t be locked inside one giant workflow. The goal is to make img2threejs modular: use what you need, replace what you don’t, build your own workflow, and eventually your own plugins. I believe this is the right foundation for img2threejs to grow from an open-source project into an ecosystem. ⁠Read the new Roadmap → / 02 / Playground The simplest way to use img2threejs: upload an image → wait ~10 minutes → get your 3D asset. Export to different 3D formats or directly to pure Three.js code. / 03 / Studio This is for those who want full control. Watch img2threejs build in real time: Nano → Texture → Rigging → Animation → VFX/SFX → Export. Configure, inspect, and control the workflow instead of only receiving the final result. / 04 / A new Landing Page Cleaner, easier, and actually more useful. You’ll be able to copy prompts directly, explore examples, and export assets for Blender or your own workflow. I’m trying to wrap up as much of the annoying work as possible animation, rigging, textures, exporters, so you can focus on building. / 05 / Official YouTube Many of you have asked how these showcases are actually built, and X simply isn’t the right place for a 20–40 minute walkthrough haha. I’ll start sharing full step-by-step workflows, experiments, tutorials, and project updates here: ⁠YouTube → And one more thing… Tomorrow at 13:00 UTC - img2threejs × GPT-6 Astra. A new exclusive showcase is coming. I’ve spent enough time with img2threejs that I thought it was getting difficult to surprise myself anymore, but this one did. To everyone who tested img2threejs, contributed code, reported bugs, shared feedback, created something with it, sponsored us, became a partner, or simply told me to keep going - thank you, truly. A lot of what you see in this new direction comes directly from those conversations. We’re still very early, and there’s still a ridiculous amount of work ahead, but I’m really excited about where this is going. If you’re building in AI, 3D, graphics, or creative tools and this direction makes sense to you, I’d love to connect. Maybe we can build something together haha. Community: Support: Same vision. Bigger possibilities. — Nick

I turned Raz from @ArenaofValor into an animated 3D character running directly in Three.js. 🔥🥊 This is my latest showcase for img2threejs v1.5.2 - and the new version is officially out. Image → TripoAI or Hyper3D → Rigging → Animation → VFX → Three.js. For this demo, I wanted to push beyond simply reconstructing a character. Raz needed to actually FEEL like Raz: explosive punches, fast movement, combat animation, and the fiery VFX around his attacks. The entire workflow was built through the img2threejs pipeline, turning visual references into a character that can be controlled, animated, and extended directly in code. And the best part: img2threejs v1.5.2 is available now. You can try the same pipeline yourself and see how far you can push AI-assisted 3D character creation for Three.js and WebGL. Open source. Built for experimentation. GitHub 👇 #ThreeJS# #img2threejs# #AI3D# #ArenaOfValor#

I’ve wanted to build this for a while. Lee Sin was my favorite champion back in middle school. 😂 So when I started pushing img2threejs 1.5.1 beyond just generating 3D models, he was the obvious choice. Image → 3D → Animation → VFX → Three.js. This is what 1.5.1 can do now. Opensource: And I’m already working on 1.5.2, a much cleaner pipeline for creating animations and VFX like this without all the manual work. Kind of cool that the champion I loved as a kid is now helping me build my tools.

Another Pure Three.js experiment with img2threejs 1.5.1. Showcase: From image → Pure Three.js in just ~30 minutes. This is the current 1.5.1 flow, and it’s getting really fast. You can simply use it to turn an image into a model you like almost instantly. Powered by Codex Luna Max. 🚀 1.5.2 is coming next week, around Monday or Tuesday, with a major focus on improving animation. I’m still thinking about the best time to post the release. What time do you think works best? #ThreeJS# #3D# #img2threejs#

Three.js community, another experiment for you ⚔️ I tried Hyper3D × img2threejs this time - using generated 3D as a spatial reference, then reconstructing it into procedural Three.js geometry. No GLB at runtime. Just programmable geometry 👀 And if you want to try the same workflow, Hyper3D gave us a community code: IMGRodinGen25 → 1 month Creator Plan for $1 🎁 Would love to see what you build with it!

Another experiment ⚔️ This time, I tried combining Hyper3D × img2threejs - and honestly, the result isn’t bad at all 👀 Image → Hyper3D → img2threejs → Procedural Three.js Showcase: This Dual-Sword Warrior is reconstructed into 31 procedural parts - no GLB, no image, and no external mesh fetched at runtime. Hyper3D provides the 3D spatial reference. img2threejs turns it into programmable geometry. And… I’ve also been working on something else: Animation for these characters. 👈 You’ll see what that looks like very soon. img2threejs v1.5.2 is coming. ⚔️🔥

A little gift for the Three.js community 🎁 Thanks to Tripo AI, you can get 60% OFF your first month: THREEJSPRO - Pro THREEJSMAX - Max Only 200 redemptions per code, so grab one if you want to try the new 3D → Three.js workflow 🚀

img2threejs 1.5.1 is out 🚀 Checkout: This release introduces a new approach to image → 3D → Three.js. Image → Tripo → img2threejs → reconstructed geometry → Three.js Instead of using the generated 3D model as the final asset, img2threejs uses it as a reference to rebuild the character into a representation that can be generated and rendered directly in Three.js. No Blender workflow, and no GLB required as the final output. The animation in the demo is still a bit rough because this release is focused on reconstruction, not animation. I’ll be improving that in the next release. Still experimental, but this is a big step toward where I want img2threejs to go.

A little sneak peek for the Three.js community This model was generated from a reference image with just ONE prompt. With img2threejs v1.5.1, we’re pushing much closer to the original reference while keeping the result structured, editable, riggable, and controllable in Three.js. And this is only a small preview. 😆 What do you think? How close are we getting? 👇