CPU3DAI tools for 3D, video and audio

ComfyUI v0.35.0: WAN3-Prime, Omni 1.1 support and new 3D nodes

On 09.09.2026, ComfyUI v0.35.0 was released. The release includes support for new partner models, 3D nodes for working with meshes, and several fixes related to memory and video.

What it means

For those running models locally, two areas are of primary interest. First, support has been added for WAN3-Prime, a new model in the WAN lineup, as well as Google's Omni 1.1 model. The developer does not specify VRAM requirements or the size of the files to download, but the appearance of support in ComfyUI means these models can be used in familiar pipelines without external wrappers.

Second, the File3DToMesh node has appeared, which parses GLB, GLTF, OBJ and STL into MESH. This simplifies working with 3D assets inside ComfyUI: a file can be loaded directly, without prior conversion in third-party tools. VideoTrim and VideoCrop nodes with VIDEO_EDIT widgets have also been added — basic video trimming operations are now performed inside the graph.

Among the fixes, worth noting is accounting for the container's cgroup memory limit instead of the host's RAM size — this matters for those running ComfyUI in Docker or other containerized environments, where the actually available memory may be less than that of the machine. HEVC remuxing in mp4/mov has also been fixed.

ComfyUI v0.35.0 — what's new New models W WAN3-Prime in the WAN line O Omni 1.1 from Google 3D nodes File3DToMesh GLB, GLTF, OBJ, STL to MESH directly no conversion Video nodes VideoTrim trimming in the graph VideoCrop VIDEO_EDIT widgets Fixes M cgroup memory limit instead of host RAM D Docker and containers available memory H HEVC remux to mp4/mov Cumulative update: models, 3D nodes, targeted fixes
How the method works. The diagram was drawn based on this news note.

The update is cumulative in nature: new partner integrations, 3D nodes and targeted fixes. You can follow news about tools for running on your own hardware in the CPU3D news feed.

See also