RovibesROVIBES // ROBLOX INFO
ENJA
Live
DevForum5 months ago

[Early Preview] Generate 3D Models Segmented Into Multiple Parts

Summary

Roblox Studio's Assistant now supports generating 3D models that are automatically segmented into individual parts, allowing for easier customization and interactivity. This early preview feature supports up to eight parts per model and introduces image-to-mesh capabilities for the generation process.

UpdateFeatureHype
Read original article →
+75Vibe

Vibe score

-100 to +100

Community reactions

DevForum
→ Neutral0 eng↗

I used roblox AI but it was alot of attempts to get it right, the exact copy of the model is мт-т

X
→ Neutral3 eng↗

We just shipped an Early Preview feature in Studio that lets Roblox creators generate 3D models segmented into multiple parts! Give it a try and let us know what you think. More to come.

DevForum
→ Neutral0 eng↗

This topic was automatically opened after 10 minutes.

DevForum
→ Neutral0 eng↗

Sooo.. why did we exactly need this again?

DevForum
→ Neutral0 eng↗

This is a great update, I can finally animate guns and cars that have multiple parts like tyres or magazines using generated models

DevForum
→ Neutral0 eng↗

…or you could just do it yourself. You know, with blender or any other suitable modelling application

DevForum
→ Neutral0 eng↗

Great update! Keep it up!! Cant wait to use this.

DevForum
→ Neutral0 eng↗

You can of course do it with a DCC tool, and the idea is not to replace this. You should use both DCC tools and gen AI depending on your goals: DCC tools are great when you want precision, for example with final assets. But generation with segmentation is super useful when you just want to test and

DevForum
→ Neutral0 eng↗

FINALLY, image prompts. after all of this time

DevForum
→ Neutral0 eng↗

Happy to see image prompts & segmenting into multiple parts, but concerned about triangle counts of these generated models now that they’re not bound to a tri limit How much do the models try to optimize on triangles?

DevForum
→ Neutral0 eng↗

You can still set a limit for the number of triangles for the overall mesh being generated.

DevForum
→ Neutral0 eng↗

NickyLarson: AI will never replace humans in the art of horse decimation (or any other art)

DevForum
→ Neutral0 eng↗

Yes! Image prompts are key and we’re continuing to explore how we can further enable creative control. More to come.

DevForum
→ Neutral0 eng↗

I do not have time nor the need when I have such tools available They probably told the inventors of the printing press to just write it out by hand.. and look what happened.

DevForum
→ Neutral0 eng↗

would be nice to have multiple image input, and an upgrade to the texturing

DevForum
→ Neutral0 eng↗

We agree Mind sharing more about how you would use multiple image input?

DevForum
→ Neutral0 eng↗

A singular image can only define one side of an object; Other sides are left for interpretation to the AI. Multiple images let you define how the object should look all-around

DevForum
→ Neutral0 eng↗

Why use a calculator when you could just do it in your head or any other thinking application

DevForum
→ Neutral0 eng↗

this is cool if you don’t like it then don’t use it very simple

DevForum
→ Neutral0 eng↗

Totally not sure why developers are hating using AI, I am using Claude Opus 4.7-8 and the projects I have done are amazing, it helped me create A-chassis tank tracks system with great performant results

DevForum
→ Neutral0 eng↗

That’s really cool! May I ask how you created the tank?