Community tips for MOSS: Turbo Lora settings, samplers, and SLG tuning
mikemend · reddit · 2026-10-24
A compiled set of practical tips for getting better images out of MOSS:
- Models: Base is more flexible, SFT more rigid.
- Loras: Turbo/DMD Loras come in versions for Base, SFT and Kromoss — grab them all (8-step versions on HF at lodestones/models); they combine well but don't work with every sampler.
- Samplers: beyond Euler/Simple, try ersde and bongtangent.
- Settings: Turbo/DMD Lora strength around 0.7 often works better; cfg 1 works in Turbo mode (author uses 1.5–1.7, up to 3), while non-Turbo benefits from cfg 3–5.
- Secret sauce: tuning the Uncond skip (SLG) node improves images — only with cfg > 1; endpercent 0.5 or 1, singlelayers 22 / 22,24 / 22,24,26.
- Extras: RGB Smart Seed Variance node adds variety; Image Sharpen from KJ-nodes (highpass/strength 0.5) gives slight sharpening.
More from Multimodal
- Midjourney sref code 1645312354 yields charming watercolor Mario-style art — sergeantsref · 2026-10-24
- Claude one-shots an 89-shot 2051 vlog with a full automated video pipeline — TuneFickle1772 · 2026-10-24
- ComfyUI's KSampler ignores denoise for step count — webui migrants hit workflow gotcha — fabie_flower · 2026-10-24
- Creator uses GPT via MCP to drive Runway SEEDANCE 2.5 for film-like video — aziz4ai · 2026-10-24
- "She said it was safe": AI-generated meerkat skydiving video delights Reddit — EnaMania · 2026-10-24
- Just Eat's AI food photos turned tiger prawns into baby tigers — a linguistics lesson in why — aakashgupta · 2026-10-24