models/DMD2 Speed LoRA [SDXL, Pony, Illustrious] - dmd2_sdxl_4step_lora_fp16

DMD2 Speed LoRA [SDXL, Pony, Illustrious] - dmd2_sdxl_4step_lora_fp16

9/28/2025

9:19:56 AM

Related Keywords & Tags

6tz,distribution matching distillation,dmd2,dmd2_sdxl_4step_lora,dmd2_sdxl_4step_lora_fp16,fast image synthesis,hyper,lcm,lora,sdxl,sdxl 1.0,speed,tool,turbo

A large futuristic tower with a wide top and mechanical gears at its base, surrounded by flowering pink trees and set against a blue sky with clouds.

Silhouette of a girl standing in a flowering meadow at sunset with a surreal twisted shadow swirling upward into a peach and mauve sky

A dark silhouette of a young woman with short hair in a high-contrast scarlet and crimson workshop illuminated by tenebrism lighting and deep shadows.

Recommended Negative Prompts

NSFW, naked, nude, watermark, text, simple background

Recommended Parameters

samplers

LCM, lcm

steps

cfg

1.4 - 1.8

other models

Checkpoints\Checkpoints\SDXL\Unsorted\juggernautXL_juggXIByRundiffusion.safetensors (33e58e8668)

Tips

Load the LoRA like any other and start with a weight of 0.7, adjusting as preferred.

Use with any SDXL or similar checkpoint for best results.

Merge the LoRA into a checkpoint to incorporate it directly and avoid separate loading.

If images appear washed out or bland, increase the CFG value.

The LoRA allows fewer steps for generation with better results but does not improve speed itself.

Version Highlights

dmd2_sdxl_4step_lora_fp16

This is the official DMD2 implementation LoRA from the Improved Distribution Matching Distillation for Fast Image Synthesis paper.

dmd2_sdxl_4step_lora

Original model at Huggingface tianweiy/DMD2.

Recommended Settings

Steps: 8

LoRA weight: 0.7

Sampler: LCM

Scheduler: Normal, Simple

CFG: 1.4 - 1.8

I use similar settings on both a low resolution base image, and the upscaled refinement pass (hires fix).

Usage

Load the LoRA like any other. I recommend a weight of 0.7, but feel free to play around.

You should be able to use it with any SDXL or similar checkpoint. Use the settings above.

You can also merge this into a checkpoint and save that, to include the result of the LoRA into the checkpoint directly. This means you no longer need to load this as a LoRA, and you can still use it with LCM and low steps for good result.

The LoRA itself doesn't add any speed improvement, it just allows you to generate at fewer steps, with better result.

If the results are washed out and bland, try increasing the CFG.

Licence

Creative Commons Attribution Non Commercial 4.0

I am not the creator of the model, just uploading the model here to make it easier to download and reference in creations. Full credit goes to Tianwei Yin.