# Hugging Face — SDXL in 4 steps with Latent Consistency LoRAs

- Company: Hugging Face (huggingface.co)
- Announced: 2023-11-09
- Category: open-weight-release
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://huggingface.co/blog/lcm_lora
- Record: https://forck.live/items/1973-sdxl-in-4-steps-with-latent-consistency-loras
- Subject: Platform
- Open weights: yes
- Models affected: SDXL, Stable Diffusion

Hugging Face announces LCM LoRAs, a method to make Stable Diffusion and SDXL generate images in 4 steps instead of 25-50, by training small LoRA adapters instead of full model distillation. The LoRAs are released on the Hub and can be applied to any fine-tuned SDXL model.

## Evidence

Verbatim from https://huggingface.co/blog/lcm_lora:

> We are delighted to announce a new method that can essentially make Stable Diffusion and SDXL faster, as if they had been distilled using the LCM process!

---

Record: https://forck.live/items/1973-sdxl-in-4-steps-with-latent-consistency-loras
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md