Turn one image into 3D splats, GLB meshes, and orbit previews.

Who it's for: creators who want this pipeline in ComfyUI without assembling nodes from scratch. Not for: one-click results with zero tuning - you still choose inputs, prompts, and settings.

Open preloaded workflow on RunComfy

Open preloaded workflow on RunComfy (browser)

Why RunComfy first
- Fewer missing-node surprises - run the graph in a managed environment before you mirror it locally.
- Quick GPU tryout - useful if your local VRAM or install time is the bottleneck.
- Matches the published JSON - the zip follows the same runnable workflow you can open on RunComfy.

When downloading for local ComfyUI makes sense - you want full control over models on disk, batch scripting, or offline runs.

How to use (local ComfyUI)
1. Load inputs (images/video/audio) in the marked loader nodes.
2. Set prompts, resolution, and seeds; start with a short test run.
3. Export from the Save / Write nodes shown in the graph.

Expectations - First run may pull large weights; cloud runs may require a free RunComfy account.

Overview

Turn your object image into a 3D Gaussian splat with Idol. Inspect it in an orbit-preview video. Export SPZ splats or a converted GLB mesh. Explore 3D concepts from multiple angles. Expect inferred details on unseen sides. Clean up the mesh before production use.

Important nodes:

Key nodes in Comfyui Idol ComfyUI Image to 3D workflow

UNETLoader (#16)

Loads the Idol UNet weights that specialize TripoSplat for stylized characters and props. If you need a baseline comparison, you can swap to the stock TripoSplat UNet from its repository. Keep the UNet, CLIP vision model, and VAEs from the same release family for stable behavior. Reference weights: Hugging Face: Comfy-Org/Idol and GitHub: VAST-AI-Research/TripoSplat.

KSampler (#6)

Drives the denoising schedule that produces the latent 3D representation. The most impactful controls are steps, sampler_name, scheduler, cfg, and seed. Increase steps for fidelity, adjust CFG to balance adherence to conditioning versus smoothness, and use the seed for reproducibility across runs. Avoid extreme CFG with very low steps, which can introduce unstable geometry.

VAEDecodeTripoSplat (#55)

Turns the final latent into a Gaussian splat. The decisive parameter is num_gaussians, which sets point count and impacts both detail and memory use. Use moderate counts for previews and higher counts for final SPZ exports. Very large counts increase render cost and do not invent new detail beyond what the model inferred.

ComfySwitchNode (#35) Switch: Mask Source

Chooses between the BiRefNet mask and any alpha or mask you provide with LoadImage. Use the model-derived mask when the input has a busy backdrop. If you supply a clean alpha, switch to your mask to preserve thin structures.

TripoSplatSamplingPreview (#97)

Provides a quick, low-cost preview of the sampling orientation and rough geometry. Enable this when first testing a subject or when changing checkpoints. It helps you verify framing and scale before committing to a full decode.

RenderSplat (#75)

Renders the orbit frames used for the preview video. The key controls are image width and height, frames, splat_scale, headlight_shading, and opacity_threshold. Keep an eye on splat scale and opacity threshold together, which govern how soft or tight the point blending appears.

…

Notes

Idol ComfyUI Image to 3D | Gaussian Splats, SPZ + GLB - see RunComfy page for the latest node requirements.