Last Week in Multimodal AI: PixText, MOSS Edit in ComfyUI, Anima refs and Hy-Nomas port

Vast_Yak_4147 · reddit · 2026-10-30

The Oct 22-28 roundup of local image/video releases with code and weights:

PixText: box-select English/Chinese text repair, 4-step MeanFlow, Apache 2.0, ComfyUI workflows, requires NVIDIA CUDA.
MOSS Edit: instruction-based editing now natively merged into ComfyUI 0.43.0.
Anima Reference Adapter: character/style reference without per-character training; noncommercial weights.
Hy-Nomas native ComfyUI port: segment chaining with remembered history and timed camera control; author notes low-VRAM claims used an RTX 4090, not actual 12/16GB cards.
ZPix 1.1.1: selection-based inpainting retaining unselected pixels, optional patch-level hi-res finishing.
Zip Ties: MIT extension bundling workflow wires into ribbons without changing execution.
CozyClay 3.2: block actors/cameras with export for video models, adds physics checks, AGPL.

Curator notes results weren't independently reproduced; licenses and hardware needs vary.

More from Multimodal