ComfyUI vs Typito

ComfyUI is the node-based generative workflow tool, and its blog reads as a running log of what has been absorbed into core. Video has been the loudest lane — Wan 3.0's 30-second native generation, Gemini Omni 1.1 Flash with 4K output and conversational editing — but 3D has been building steadily since Tripo 3.1 in May and native Gaussian Splats in June. Alongside the model work, Comfy became an official reseller of MiniMax H3 commercial licences and open-sourced its local MCP server.

The pattern is consolidation: capabilities that used to require custom nodes, compiled CUDA extensions or non-commercial licences keep moving into core with a supported pipeline around them. The Trellis.2 and Pixal3D integration is the clearest statement of it, arriving with rebuilt Load, Preview and Save 3D nodes, mesh post-processing and PBR texturing rather than just model weights. The licensing work runs in parallel and matters as much — Comfy is making commercial use of these models straightforward, not only technically possible.

3D now has a core pipeline the way video does, so expect the next 3D models to arrive as ordinary integrations into that pipeline rather than as their own announcements, and expect the licensing-reseller pattern to extend to whichever 3D model needs it.