Turn a single 2D image into a fully interactive, animated 3D character - running directly in your…
@NickDevFE3K followers
49,250 views
55 Comments
Comments from the original post on X (Twitter)

@NickDevFE
Here's prompt I used:
25 likes

@GamefxAI
Ignore the haters keep experimenting!
3 likes

@NickDevFE
Appreciate you brother! 🙏 I’m definitely not stopping, just going to keep experimenting, building, and pushing this thing further.
2 likes

@FabioAngela79
Cool! How did it rig the model, do you use an already made skeleton? And what about weights? Defo it has some problem (tested the Body Shot Taken) but for sure it's improving! (in my toolings I'm trying UniRig and Make-It-Animatable and they both can't produce perfect weighting that I can properly animate with Kimodo)
3 likes

@NickDevFE
Thanks for testing it and for the detailed feedback! 🙏 For the current experiment, I’m using a reference GLB to extract the skeleton/specs and then generating the rigging setup in code. The weights are also generated as part of the pipeline, I’m not using a manually authored skeleton + hand-painted weights. And yeah, Body Shot Taken definitely exposes some weighting/rigging issues. That’s actually one of the areas I’m actively experimenting with right now. UniRig and Make-It-Animatable are interesting comparisons. Perfect weighting that works reliably across arbitrary characters is still a really hard problem. Animation/rigging is still on my roadmap, so this is very much an early experiment. I’m hoping to push this much further in the next iterations. 🔥
6 likes

@KevinBlumson
@Andrew_Blumson pretty good!
2 likes

@NickDevFE
@Andrew_Blumson Thank you! 🙌 Really appreciate it!
2 likes

@AbraTabia
Pipeline is impressive. The thing that decides whether it lands for players is what the GLB weighs after reconstruction. Browser games live or die on time to first interaction, and a generated mesh will happily out-detail what a phone can download. Is there a decimation step?
2 likes

@NickDevFE
Exactly. And you can absolutely control the resolution too. In some of my earlier showcases, I added High / Medium / Low options to reduce the triangle count depending on the target. For a real production setup, you could even make it adaptive based on the device e.g. mobile → Low, desktop → Medium/High, or inspect the device specs and choose automatically. So the pipeline can be tuned for the actual target hardware rather than always shipping the highest-detail mesh. 🔥
1 like

@minkrov
Image to playable 3D in the browser is huge for web games. Building one with Remie right now to ship on Poki/Crazy Games.
2 likes

@NickDevFE
That’s awesome! Image → playable 3D is exactly the direction I’m pushing img2threejs toward. Would love to see what you’re building for Poki/Crazy Games when it’s ready. 🚀
1 like

@internetlabsai
Turning a single 2D image into an animated, fully rigged 3D boxer with physics sweat particles running on pure Three.js without ever opening Blender is pure dark magic
2 likes

@NickDevFE
Thanks, Appreciate it! The wild part is that the whole thing is generated directly as Three.js code, no Blender in the workflow. Still a lot to improve, but it’s getting there.
1 like

@zenkiauph
those boxing animations hahahahahahahahahahahahaha holy dog shit
1 like

@NickDevFE
LMAOOO 😂😂 I know right?? They’re absolutely unhinged 🔥

@zenkiauph
if unhinged means unusable and useless yeah very unhinged bro 🔥🔥🔥 elite rigging bro 💩
1 like

@NickDevFE
LMAO fair roast 😂 v1.5 is mainly about creating the character, animation is still on the roadmap. I’m just experimenting with it early. According to the roadmap, I wasn’t even supposed to be doing this yet Still plenty of room to cook, its 1.8 🔥

@zenkiauph
that model is horrible for 100k, if you are going to showcase this webapp, don't use a boxing character of all characters 🤣 use a walking old man or something just wasting your claude credits brother🤣nobody in their right mind is using that model and animation brother, sloppy
1 like

@NickDevFE
I get what you mean, brother 😂 But every successful project starts by finding its failure points. That’s exactly why I don’t hide my older showcases, even when they’re fucking rough. I use these showcases to document the process of improving img2threejs. Every showcase is an iteration, and every iteration gets better. Some people look at the improvement. Others only look at what’s still broken. Either way, I’m gonna keep cooking. 🙏

@Dedbwur
Skipping Blender entirely for a pure TypeScript output is a massive workflow win for web developers.
1 like

















