Sampling Method
The technique the AI uses to turn noise into your image — each sampler has its own personality.
The sampling method (or sampler) is the technique the AI uses to remove noise at each step of generation. Different samplers are like artists with different working styles — same brief, noticeably different results, even with an identical Prompt and seed.
Quick guide to PixAI's samplers
Euler a — the PixAI default. PixAI's in-house models are optimized for it. Start here.
The "a" (ancestral) family —
Euler a,DPM2 a Karras,DPM++ 2S a Karras: inject a bit of randomness at every step, so results keep evolving. More creative surprise, less reproducibility — even the same seed can differ slightly.Euler / LMS / Heun — the classics. Euler is simple and fast for quick iteration; Heun trades speed for accuracy; LMS handles smooth gradients (hair shading, skin tones) nicely.
DDIM — the batch workhorse. Produces good quality in fewer steps; great for generating lots of candidates.
DPM++ 2M Karras & friends — the detail lovers. Excel at intricate outfits, accessories, and complex backgrounds. The "SDE" variants add controlled randomness for dynamic poses and flowing hair.
The one rule that matters
Check the model's description page. Not every community model supports every sampler — a mismatch can output pure noise. Model creators usually list their recommended sampler; start there, then experiment. (All of PixAI's in-house models work with every sampler.)