Updated: Oct 30, 2026
style
Main characteristics of this LoRA are as follows:
· Reversal film / slide film style, with strong contrast and rich, intense color saturation.
· Improved lighting and shadow effects, especially extremely powerful and expressive sunlight rendering.
· Based on personal aesthetic preference, female characters tend toward large breasts, slim waist, and wide hips.
Important: This LoRA has a very strong style with distinct personal preferences. In other words, it cannot guarantee diversity in body types etc. (You can achieve more variety by using detailed and specific prompts.)
Info for MOSS-Image-Base v1.0
This LoRA was trained on MOSS-Image-base and can be used with lodestones/MOSS-Image-Base-8-steps-dmd. In theory, it should also work with the SFT model, but this has not been tested.
Recommended Settings:
Strength: 1
Sampler: euler_sgm_uniform
Info for Flux.3 Klein v1.0
This Lora supports both base and step distilled model. NSFW supported.
Recommended Setting:
strength: 1
sampler: euler_simple
License NOTE：
This LoRA is released under the FLUX Kommunity License v.1.0: non-commercial use, plus commercial output generation by Qualifying Users with less than US$5M in gross annualized revenue, including affiliates, subject to the license terms. For other uses, see commercial licensing.
Info for QwenImage 2.1 v1.0
Recommended strength: 1.0
Recommended sampler & scheduler: euler_simple
NOTE：
License
This LoRA is a derivative work based on Qwen-Image-2.1. It is released under the Qwen Research License Agreement. Commercial use is not permitted without the appropriate commercial license from Qwen. Please refer to the original Qwen-Image-2.1 license: https://huggingface.co/Qwen/Qwen-Image-2.1/blob/main/LICENSE
Info for Krea2 Turbo v1.0
Recommended strength: 0.5–1.0, 0.75 is the best in my test.
Recommended sampler & scheduler: er_sde_simple
Info for ERNIE-image (and turbo) v1.1
This LoRA is compatible with ERNIE-image and its distilled model ERNIE-image-turbo.
For the ERNIE-image model, the default recommended settings are CFG 4.0 / steps 25+.
It can also be used in combination with the Turbo LoRA (GuangyuanSD/ErnieTurbo_extracted_lora · Hugging Face) at CFG 1.0 / steps 8.
For the ERNIE-image-turbo model, it can be used directly.
In addition, this LoRA can significantly reduce the yellowish cast and rough artifacts of the ERNIE-image-turbo model, producing images with better white balance and finer, more detailed quality.
Recommended strength: 1.0
Info for z-image-turbo v2.0
This is an improved version of v1.0, with the main changes as follows:
· Overall tone is brighter, with more details preserved in the shadows
· Reduced some unreasonable light and shadow effects
· Contrast is not as strong/intense as in v1.0
Rather than calling it an “upgrade,” it’s more accurate to say it’s a variant with a different color tone and style. Users can choose between the two versions based on their personal preference.
Recommended strength: 0.7–1.0
Recommended sampler & scheduler: euler_capitanZIT (https://github.com/capitan01R/ComfyUI-CapitanZiT-Scheduler)
Info for z-image-turbo v1.0
Recommended strength: 0.5–1.0
Recommended sampler & scheduler: euler_capitanZIT (https://github.com/capitan01R/ComfyUI-CapitanZiT-Scheduler)
Info for z-image v1.0
This is a version retrained based on the z-image model.
Recommended strength: 1.0
Recommended sampler & scheduler: res_multistep beta 25steps