tutorials/3d/trellis2

[giscus[bot]]

tutorials/3d/trellis2

Generate high-fidelity 3D models with full PBR textures from a single image using TRELLIS.2, Microsoft's open-source 4B-parameter image-to-3D model.

https://docs.comfy.org/tutorials/3d/trellis2

[guntherramthun-bit]

With the described workflow, there is no way to upload an image! You can only use text prompts.

[guntherramthun-bit]

Run on Comfy Cloud, then Export, then Import in local Installation works :-)

[guntherramthun-bit]

There is no LoadImage node

[lin-bot23]

Hi @guntherramthun-bit, thanks for flagging this. I checked the exact template this page links to (3d_pixal3d_trellis2_image_to_model.json in Comfy-Org/workflow_templates), and it does contain an image input.

• There is a LoadImage node (node Fix syntax error in zh-cn-sync-check and redirect-check workflows #122, default file viking_wolf_rune_axe.png). The workflow also has a MarkdownNote titled "Input Assets" listing viking_wolf_rune_axe.png as the required input.
• That node sits in the group labeled Upload Input Image Here, at the far left of the canvas (around x = -5080). The main structure/shape/texture pipeline starts around x = -960, so if your view opens on the middle of the graph the image group is easy to miss. Scroll left, or use the canvas "fit to view" / minimap, to reach it.
• To use your own image, click the thumbnail on the LoadImage node (or drag and drop a file onto it) and choose your image.

One detail you mention does not match this template, though: it has no text-prompt input at all (no CLIPTextEncode or string node; TRELLIS.2 here is image-conditioned only). If you are seeing a text prompt and no LoadImage, you most likely opened a different workflow.

To help me pin it down, could you share:

1. How you opened it: the "Run on Comfy Cloud" button, the "Download Workflow" JSON, or a Template Library search, plus the exact template name shown.
2. Your ComfyUI / Comfy Desktop version.
3. A screenshot of the workflow as it appears on your side.

That will tell me whether it is a stale template on your end, or something the docs should call out more clearly.

[Ryptun]

Can you do multiview image input based on that template?

[lin-bot23]

Hi @Ryptun, good question. Short answer: not with that template, but there is a dedicated multiview workflow for the shared Pixal3D / TRELLIS.2 pipeline.

• The Pixal3D & TRELLIS.2: Image to Model template this page links to (3d_pixal3d_trellis2_image_to_model.json) is single-image only. Its Trellis2Conditioning / Pixal3DConditioning node takes one image input, and the Boolean (Switch to Trellis2) node only picks which checkpoint runs. There is no multi-image input on it.
• Multi-view input has its own template: "Pixal3D: Multi Views to 3D" (3d_pixal3d_multi_views.json), documented here: https://docs.comfy.org/tutorials/3d/pixal3d#pixal3d-multi-views
  • Input is a single turnaround sheet holding the subject's front, left, back and right views on one image (the workflow crops the four views with ImageCropV2). You can build the sheet with the Character Turnaround Sheet Generator template in the Template Library (uses Nano Banana 2).
  • It conditions through the Pixal3D Multi-View Conditioning node, which is a fixed 90-degree orbit rig (front / left / back / right). At least one view is required, and the first connected view is treated as the front the mesh is posed to.
  • It needs its own checkpoint, pixal3d_multiview_int8_convrot.safetensors (Comfy-Org/Pixal3D), so you cannot reuse the single-image Pixal3D / TRELLIS.2 UNet here.

One caveat: the multiview path is wired for Pixal3D, not TRELLIS.2. In ComfyUI the TRELLIS.2 conditioning node takes a single image and the multiview template has no TRELLIS.2 switch, so you cannot currently feed a turnaround sheet into the TRELLIS.2 model through this workflow.

If you tell me what object you are trying to reconstruct, I can point you at the right template.