OpenCartridge adds Kimodo text-to-VRMA and multi-image UniForge3D for local game assets

OpenCartridge, a local-first suite for building game art, added two substantial 3D production cartridges on November 7, 2026. A new Kimodo worker converts written motion instructions into a VRM animation file (.vrma), while UniForge3D lets creators choose one or more canvas image layers to reconstruct a textured GLB asset. These are documented source changes, not evidence that a standalone new commercial model launched that day. The feature additions are visible in the Kimodo integration commit and UniForge3D integration commit.

Kimodo: motion prompts become VRMA clips

The November 7 commit adds a Kimodo cartridge to the service registry at port 8013, with a text prompt input and .vrma output. Its description gives actions such as walking, waving or dancing; multiple sentences can form sequential motion segments joined with blended transitions. The worker describes mapping model joints to VRM humanoid bones and exporting compatible animation. It is intended for a VRM avatar pipeline, not an automatically assembled playable game.

For creators, the practical workflow is: prepare a VRM humanoid, describe an action in text, generate a VRMA clip, import or preview the motion in a compatible avatar viewer, then check foot contact, timing and transitions. The published integration code is available for inspection; this report does not claim to have run the model.

UniForge3D: multiple views into one GLB

A separate November 7 integration adds UniForge3D, accepting one or several selected image layers. The source change extends the input-binding rules for optional numbered image slots, so an artist can supply multiple views in one request. The documented cartridge aims to produce a textured GLB object with PBR material, reconstructing unseen sides where possible. The advantage is keeping visual references within a single canvas-based workflow rather than manually exchanging image files between tools.

Availability and limits

OpenCartridge runs locally through Docker and GPU-backed workers; the README says current worker options require an NVIDIA GPU with approximately 24 GB of VRAM. The software itself can be self-hosted without a cloud account, though local compute and separate model conditions still apply. Kimodo and UniForge3D are integrations in the OpenCartridge repository, not independently validated end-to-end results. Review the repository setup guide and original source commits before using production assets.

Sources: Kimodo source change, Nov 7 · UniForge3D source change, Nov 7 · README & setup.