MOSS-Image ecosystem roundup: LoRA training tips, sigma schedulers, ONNX port

optimisticalish · reddit · 2026-10-25

A daily roundup of MOSS-Image community resources from October 25:

Character LoRA training advice plus an Automagic V4 settings config for ostris/ai-toolkit.
An autocaptioner that captions LoRA training sets locally via llama.cpp and LLMs.
Three general-purpose sigma schedulers for ComfyUI MOSS-Image workflows (Shift, Curve, Sigma).
Style LoRAs: a soft-3D children's illustration look (useful for icon sets) and medieval parchment/mural art.
An ONNX implementation of MOSS-Image on Hugging Face.

The post links to the previous eight daily roundups, making it a running tracker of the model's ecosystem.

More from Multimodal