ComfyUI team open-sources MOSS Image LoRa for 4MP+ high-res generation
Crazy-Repeat-2006 · reddit · 2026-10-23
The ComfyUI team released a LoHa adapter, 'MOSS Image MAX ULTRA hires,' trained on top of MOSS-Image-Base to improve generation quality at 4MP and above, using a mix of heavy anime data and 3.5MP–10MP photographs. Weights are open-sourced on Hugging Face for direct use in ComfyUI workflows.
More from Multimodal
- Moss image model + HiToon LoRA showcase: fast generation with strong prompt adherence — Putrid-Dark2076 · 2026-10-23
- Gemini 4's animated SVG generation praised as exceptionally strong — DynamicWebPaige · 2026-10-23
- Visualizing probability paths in diffusion: straight lines vs elliptical arcs — alec_helbling · 2026-10-23
- GPT Image Stealth Models Slammed for Low-Taste Output; Users Want Finetuning — flowersslop · 2026-10-23
- Open-sourced Claude Code skill: turn product photos into Seedance video ads automatically — tomcrawshaw01 · 2026-10-23
- Cosmic Watercolor Fusion: a reusable image-gen prompt template with stellar results — LudovicCreator · 2026-10-23