ComfyUI Node
BD TRELLIS.2 Multi-Image Conditioning
Multi-image conditioning for TRELLIS.2 multi-view generation. Takes a BATCH of images + masks and encodes each with DinoV3, then concatenates the feature tokens for multi-view conditioning. This enables the model to see multiple views of an object and generate more accurate 3D shapes. Usage: 1. Batch multiple reference images (different angles) using a Batch node 2. Batch their corresponding masks 3. Connect output to BD Image to Shape or BD Shape to Textured Mesh The model attends to all image features simultaneously, similar to TRELLIS1's multi-image mode. Resolution modes: - 512: Fast, lower detail - good for shape - 1024_cascade: Balanced - 1536_cascade: Highest detail - best for texture
BD TRELLIS.2 Multi-Image Conditioning
- images
- masks
- conditioning
- preprocessed_images
◄resolution1024_cascade►
◄attn_backendflash_attn►
◄vram_modekeep_loaded►
◄include_1024true►
◄background_colorgray►
Category🧠BrainDead/TRELLIS2
Inputs (7)
| Name | Type | Default | Description |
|---|---|---|---|
| images | IMAGE | Batched images (B > 1) - use a Batch node to combine multiple views | |
| masks | MASK | Batched masks corresponding to each image | |
| resolution | COMBO | 1024_cascade | Model resolution mode |
| attn_backend | COMBO | flash_attn | Attention implementation |
| vram_mode | COMBO | keep_loaded | VRAM usage strategy |
| include_1024 | BOOLEAN | true | Extract 1024px features (needed for cascade modes) |
| background_color | COMBO | gray | Background color for preprocessing |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| conditioning | TRELLIS2_CONDITIONING | — |
| preprocessed_images | IMAGE | — |