ComfyUI-RMBG v3.2.0 Adds VRAM Unload Toggle and Cuts Startup From 15s to 0.1s
Narrow-Particular202 · reddit · 2026-10-01
The ComfyUI-RMBG background-removal/segmentation node pack shipped v3.2.0 with substantial updates.
Two new nodes
- Resolution Selector: given an aspect ratio and target megapixels, it computes and rounds width/height and returns an empty latent directly
- Image Compare View: renders two-image comparison on the canvas with a draggable wipe line, opacity blending, pixel-difference maps, and a difference mask output
Model unloading: every node that loads a large model now has an optional unloadmodel toggle that frees VRAM immediately after the node finishes.
Fixes: transformers 4.x compatibility (5.0 broke RMBG-2.0, SAM, SAM2, GroundingDINO), batch dimension mismatches, SDMatte crashes on unmasked input, new SAM3 Multiplex support with native safetensors, a fix for Apple Silicon nodes silently falling back to CPU, and startup time cut from 15s to under 0.1s.
More from Multimodal
- Free Tesseract extension brings design and motion graphics into ChatGPT/Claude — aziz4ai · 2026-10-01
- ComfyUI SeedVR2 upscaler v1.5.7 tames DiT VRAM spikes, runs TensorRT under 6GB — simmol · 2026-10-01
- Commentary: Big tech dissolved music's context, making 'Personal AI Music' possible — pixlpa · 2026-10-01
- Dev tests NVIDIA SoL-Refiner in ComfyUI: more like img2img reimagining than super-resolution — imlo2 · 2026-10-01
- HiDream-O1-Video debuts at #7 on Image-to-Video Arena with 1456 points — arena · 2026-10-01
- Higgsfield launches Genjutsu Restyle with 20+ animation styles, now via API and MCP — tisch_eins · 2026-10-01