These samplers apply the LoRA at 100% strength:
1) DPM++ 2S a — ★ Best Overall
Maximum sharpness, strong color, excellent micro-detail.
No need to lower LoRA strength.
Best: 5–(8,9,10,etc). steps, CFG 1, LoRA 1.0.
2) Euler A — ★ Cleanest Style
Extremely consistent in low steps.
Great for anime/cartoon styles.
Best: 4–(8,9,10,etc). steps, CFG 1, LoRA 1.0.
3) DPM++ 2M SDE — ★ Stable & Balanced
Nice mix of detail + smooth rendering.
Best: 6–10 steps, CFG 1, LoRA 1.0.
4) DPM++ SDE
Similar to 2M SDE but slightly softer.
Best: 6–10 steps, CFG 1, LoRA 0.6–0.9.
5) DPM Adaptive — ★ Highest Quality (but slow)
Produces extremely clean images even with 1–3 visual steps.
⚠ Slower because it performs many hidden micro-steps.
Best: 1–4 steps, CFG 1, LoRA 0.5–0.8.
🟨 GOOD (with specific settings)
6) DPM++ 2M
Works well only with CFG ≥ 1.5
Best: 6 steps, CFG 1.8–3, LoRA 0.5–0.8.
7) DPM2
Surprisingly good at low step counts.
Best: 3–4 steps, CFG 1–1.8, LoRA 0.7–1.0.
🟥 NOT RECOMMENDED / INCOMPATIBLE
These samplers do not activate the LoRA’s improvements correctly:
✗ Heun ✗ DPM2 a (ancestral) ✗ DPM++ 2M ancestral ✗ DPM++ 2S ancestral ✗ Any ancestral sampler except Euler A
Typical issues:
muddy/low color
broken detail
gray shadows
unstable composition
noisy lines
🎚 Recommended LoRA Strength
0.7–1.0 → maximum color, contrast, and detail
0.5–0.7 → extra stability on heavy base models
0.3–0.5 → subtle, soft enhancement
This LoRA does NOT require high CFG. Best CFG range: 1.0–2.5.
🔥 Recommended Step Count
Best range: 4–8 steps
Works even at 1 step (Adaptive)
8–12 if you want finer micro-details
📌 Super Short Summary
If you don’t know what to use:
⭐ DPM++ 2S a → maximum detail & contrast
⭐ Euler A → clean, stylized results
⭐ DPM Adaptive → highest quality in low steps (slower)