What is the SDXL refiner for?

The refiner is a second specialized model trained on the low-noise end of the diffusion process. The base builds the overall composition; the refiner takes over near the end to add fine detail, sharper textures, and cleaner edges. It is most noticeable on faces, hair, and intricate surfaces.

What split between base and refiner works best?

A common starting point is the base handling about 80% of the steps and the refiner the final 20%, often written as a refiner switch around 0.8. So at 40 total steps the base runs ~32 and the refiner ~8. Pushing the refiner share too high can wash out detail.

Do I have to use the refiner?

No. Many fine-tuned SDXL checkpoints and community merges produce excellent results from the base alone and the refiner adds little. The refiner is most valuable on the original Stability base models; test both for your checkpoint.

What denoise should the refiner use?

In the ensemble-of-experts setup the refiner continues the same latent, so its effective denoise is just the remaining noise fraction (around 0.2 if it starts at 80%). If you run the refiner as a separate img2img pass instead, a denoise of about 0.2–0.3 is a good range.

SDXL Base + Refiner Workflow Guide

SDXL base + refiner workflow

Stable Diffusion XL ships as two models working together: a base that establishes the composition and a refiner that polishes the final stretch of the denoising process. Splitting the work this way can sharpen faces, hair, and fine textures — provided the handoff point is set sensibly. This guide turns your total step count and switch point into the exact per-model settings.

How it works

In the recommended “ensemble of experts” setup, both models share a single latent. The base denoises from full noise up to a switch step, then the refiner continues from there to the end. The switch is usually expressed as a fraction — around 0.8 means the base does the first 80% of steps and the refiner the final 20%. Because the refiner only handles the low-noise tail, its job is detail, not structure: too large a share and it has nothing left to refine and can soften the image. This tool splits your total steps at your chosen point and reports the base steps, refiner steps, and the refiner’s effective denoise.

Tips and trade-offs

Start at an 80/20 split. It’s the most reliable balance of detail and speed for the original SDXL base + refiner pair.
Test base-only first. Many community checkpoints already bake in the refiner’s strengths — if base-only looks great, skip the second model and save time.
The refiner adds latency. Two models mean two loads and two passes; on limited VRAM the speed cost can outweigh the detail gain.
Faces and hair benefit most. Reach for the refiner on portraits and detailed close-ups, less so for flat or stylized art.