← sources

A quick MOSS-Image news round-up - 20th October 2026

source: reddit (reddit:comfyui)published 2026-10-20 13:08ingested 2026-10-21 02:40

retained text

Another quick MOSS news and goodies round-up, for those who may have missed some items. 

-> The MOSS-Image Color-Depth Control LoRA now has integrated ComfyUI support in the latest Comfy Portable, thanks to a pull/merge request by Kijai. He has also shared a workflow. We thus now have at least basic Controlnet-like functionality in MOSS-image, via depth and colour inputs. I'd add that the depth/colour renders required by the LoRA can also have poses added by using 3D-figure desktop software such as Poser 12.

https://huggingface.co/darask0/MOSS-Image-color-depth-control

https://github.com/Comfy-Org/ComfyUI/pull/17206 (includes Kijai's workflow)

https://graphixly.com/products/poser-pro (the full Poser 12 for $20, for figure poses on custom 3D figures + easy depth renders)

-> ComfyUI-MOSS-Image-Ostris-Edit. Run image-edit LoRAs with MOSS, e.g. convert anime style to realistic.

https://github.com/SurpassHR/ComfyUI-MOSS-Image-Ostris-Edit

https://civitai.com/models/3009913/moss-edit-anime2real-anime-to-realistic (anime to realistic LoRA)

-> A first proof-of-concept in using character reference images with MOSS-Image.

https://huggingface.co/pleiades7/MOSS-Image-CharRef-PoC

-> Fizgig is continuing to very rapidly update to support LoRA training for MOSS, and is also continuing to nail down some of the early problems people had. Also note that, since last mentioned, the latest Fizgig v8.7.9 has quickly tested and jettisoned yesterday's Kromoss finetune... "Kromoss is no longer offered for MOSS-Image, as the preview model or as a base to train on. As the preview model, its previews were unreliable: it's too different from MOSS-Image Base, so a LoRA trained on Base doesn't show reliably on it. As a base to train on, it overcooks".

https://github.com/shootthesound/Fizgig/releases

-> The maker of Fizgig also has a new 'Fizgig MOSS-Image SFT Look' LoRA...  a 2Gb "LoRA of the difference between MOSS-Image SFT and MOSS-Image Base. On Base it gives renders which have much of SFT's look, but without loading SFT." Sounds good, and presumably at 0.5 strength one might get the best of both?

https://huggingface.co/ShootTheSound/Fizgig-MOSS-Image-SFT-Look-LoRA

-> Advice on manually tweaking settings in Spectrum with MOSS-image, if speed is the most important thing for you.

https://www.reddit.com/r/StableDiffusion/comments/1xaby5q/moss_image_speed_boost_with_spectrum_moss_node/

-> An initial release of EntroPack's Pre-Quantized MOSS-Image models for DiffSynth-Studio.

https://huggingface.co/DiffSynth-Studio/MOSS-Image-EntroPack

-> And finally, an early MOSS-Image inpainting test experiment with a shared workflow.

https://www.reddit.com/r/comfyui/comments/1xa9cf7/i_find_moss_model_to_be_incredible_at_inpainting/

https://civitai.com/models/3010615/moss-image-inpainting-workflow

~ OLD POSTS ~

https://old.reddit.com/r/comfyui/comments/1x97q81/a_quick_mossimage_news_roundup_18th_october_2026/

https://old.reddit.com/r/comfyui/comments/1xa3yk2/a_quick_mossimage_news_roundup_19th_october_2026/

document fetched 2026-10-21 00:41 · canonical https://reddit.com/r/comfyui/comments/1xarop9/a_quick_mossimage_news_roundup_20th_october_2026