New ComfyUI node automates image upscaling with context-anchored tiles
blakeem · reddit · 2026-07-24
- A ComfyUI user released Context Anchored Tile Refine, a new node for fast, seamless image upscaling and refinement.
- The author says it was rebuilt from scratch after a fork of ComfyUIUltimateSDUpscale ran into design problems, legacy code, and a name conflict.
- The node uses a deterministic tiling method, auto-calculates tiles from a maximum output size, and bakes in recommended settings so users do not need to guess tile sizes.
- It also supports masking different regions and refining them with different models or settings in a single node.
- The author added a Tile Simulator to visualize tile placement, overlap blending, and the contextanchor mechanism that preserves coherence across tiles.
- They say they use it with Chroma for generation, Z-Image Turbo for 3x upscaling, and SAM3.1 segmentation to treat characters and backgrounds separately.
- The node is available now in the ComfyUI Node Manager, with the repo and registry links shared in the post.
More from Multimodal
- Magnific teases October 28 release: 'We're making it move' — cuenca · 2026-10-27
- Midjourney --sref demo: retro Americana Batman art series — gizakdag · 2026-10-27
- AI video is a new medium, not cinema or animation — and it's perfect for depicting dreams — Kyrannio · 2026-10-27
- Dad turns his toddler's 31-year-old quote into a song via SUNO kids' choir — NathanpmYoung · 2026-10-27
- Band's local-AI music video hits 200k views: 900 generations kept 64, 3-second clips beat face drift — Hwang0500 · 2026-10-27
- Adobe Firefly video editor walkthrough: one-sentence Quick Cut turns 32s into 20s — LudovicCreator · 2026-10-26