Schedulers
Learn about the Scheduler and Time Step Spacing options in Sogni
Last updated
Was this helpful?
Learn about the Scheduler and Time Step Spacing options in Sogni
Last updated
Was this helpful?
Was this helpful?
Schedulers control the denoising process — the sequence of steps that transforms random noise into your final image. Each scheduler has its own way of removing noise, affecting sharpness, realism, color, and speed. Time Step Spacing changes how those steps are distributed, further influencing the look and feel.
Flux / Chroma models: Euler, Euler a, DPM++ 2M.
Stable Diffusion standard, XL, XL Turbo, and LCM models:
With Fast Supernet: DPM Solver Multistep (DPM-Solver++), PNDM (Pseudo-linear multi-step), LCM (Latent Consistency Model), Euler, Euler a.
With Relaxed Supernet or on-device: DPM Solver Multistep (DPM-Solver++), PNDM (Pseudo-linear multi-step), LCM (Latent Consistency Model).
SD3 models: Discrete Flow Scheduler (SD3) only.
Euler – Simple Fastest, most stable default. Good balance of sharpness and mood. Recommended baseline for all Flux-family models.
Euler – Beta Warmer colors and slightly sharper micro-contrast. Popular with Chroma users.
Euler – Normal Similar to Simple but with a different noise decay curve. Works well for Chroma builds ≥ 0.18.
Euler – DDIM Adds extra punch to edges. Safe alternative to Simple/Beta.
DPM++ 2M – Beta Slower but the go-to for photorealism. Keeps skin clean and reduces “Flux face” artifacts.
Euler a – Simple / SGM Uniform (Experimental) Higher variation and motion blur for stylized/anime work. Requires the July 2025 ancestral-sampler patch.
DPM Solver Multistep (DPM-Solver++) – Reliable all-rounder for non-Flux models.
PNDM (Pseudo-linear multi-step) – Uses PLMS without PRK steps.
LCM (Latent Consistency Model) – Designed for LCM and SDXL-Turbo models; works best with 2–8 steps and low guidance (1–2).
Discrete Flow Scheduler (SD3) – For rectified flow–based SD3 models only.
Controls how steps are distributed:
Flux models: Simple, SGM Uniform, Beta, Normal, DDIM.
Stable Diffusion standard, XL, XL Turbo, and LCM models: Linear, Leading, Karras.
Spacing can subtly change texture, detail, and smoothness. For SDXL models, Karras is the default — but for animations with a guide image, Linear or Leading can help prevent degradation.
Need help? Ask in our Discord!
For any questions regarding the use of Sogni, tips, suggestions, bugs, sharing art, discussing AI and life, or simply hanging out... Join our Discord community! Join Sogni Discord ✨