@Novak253@mattworkman Thanks Donovan. That's roughly what we built Mixar for. It's a Blender fork, so the full DCC is there, with an AI agent layer on top that can drive the whole thing, not just generate assets. @mattworkman curious whether it clears the bar you're describing.
Hosted our first webinar today: 40% Faster Production Ready 3D.
Local models, live workflow, concept to production-ready with no cuts.
Blender artists, game devs and product viz folks showed up and kept us well past the hour with questions.
Running these weekly on Discord!
Seedance 2.5 is now live in Mixar.
• Your 3D scene is the input, so the set never drifts between shots
• Full camera control, push in, orbit, track, whatever the shot needs
• Blockout to finished render without leaving the editor
Same world, any angle.
Mixar Patch 3.1.35 is live 🚀
→ Codex integration — plug your Codex sub straight in and run the agent. No API keys, no setup.
→ OpenRouter support — 400+ models inside Mixar. Swap models, tune parameters, see what performs.
→ Auto Rig — rig any mesh through the agent, Tripo-powered. No manual armature work.
→ @ mentions — reference any object, material, or collection from your scene, so the agent locks onto exactly what you mean.
Also shipping: chat history, agent feedback ratings, plus bug fixes and stability improvements.
Update and dive in
https://t.co/sMsw1GYBUs
Love seeing V3.0 in the wild. Huge thanks to @Stefan_3D_AI for the review! Parallel agents + BYOK + a faster agent, built directly into the core. We're just getting started.
Found a very impressive open-source AI + Blender experiment.
It’s basically a Blender fork with an AI agent inside the app itself, so it can work closer to the core instead of relying on MCP or plugins.
Didn’t expect much, but in some production tasks it actually worked better and faster. You can connect any AI like Fable 5 with API key.
Baking, UVs, LODs, physics, scene generation — full test is in the video.
Respect to @mixie3D for making it open source 👇
https://t.co/kvLMrIrp9D
Fable 5 is delivering in Blender!
We benchmarked Fable, GLM, Sonnet 5, and GPT 5.5 side-by-side (raw outputs only).
While GPT 5.5 went wild with overexposed lighting, Fable 5 dropped a remarkably clean final render.
@Edgar__s@shedntcare_@AnthropicAI Honest answer: we're not selling "production-ready output." We're selling control. One editor, full pipeline, AI that speeds you up without taking the wheel. What production-ready looks like? That's your call, not ours.
@extellacreates@AnthropicAI Both are valid. But connectors break the moment the context between tools doesn't match. When AI lives inside the editor, it sees your scene, your materials, your lighting. That's what makes the output actually usable.