Red-headed girl in a blue plaid shirt and black pantyhose sits on a subway train at night, holding a smartphone and wearing earbuds, surrounded by other passengers.
A smiling girl with curly golden blonde hair and blue eyes hugging a large teddy bear while wearing blue polka dot pajamas, sunlight streaming through a window behind her.
Smiling robotic housekeeper with mechanical arms wearing a maid outfit, holding a laundry basket outdoors with fluttering white sheets and blue sky in background
Close-up portrait of a cybernetic girl with short white to blue hair, bright purple eyes, mechanical skin, and mechanical headset parts, featuring white parts with industrial decals on a white background in pencil sketch style.

Zalecane podpowiedzi

Smooth Quality, 1girl, grin

Zalecane negatywne podpowiedzi

Smooth Negative​​​

Zalecane parametry

samplers

DPM++ 2M

steps

20 - 30

cfg

2 - 5

clip skip

2

resolution

832x1216

vae

sdxlVAE_sdxlVAE.safetensors

other models

Smooth_Lighting_Enhancer (c2ada06006fd), Smooth_Booster (2e1d7277a565), SmoothMix_Illustrious (395773d6d5), betterDaysIllustriousXL_illustriousXLV01 (05369ba2dd), waiNSFWIllustrious_v90 (178fc987b2), pasanctuarySDXL_v50 (b8616b3a8f)

Zalecane parametry wysokiej rozdzielczości

upscaler

4x_foolhardy_Remacri

upscale

1.5

steps

10

denoising strength

0.2

Wskazówki

Use weight between -0.5 and 0.5 for 90% of cases.

Negative weight (-1.0 to 0.0) emphasizes shadows and dark colors; positive weight (0.0 to 1.0) emphasizes light sources and bright colors.

Low CFG (3.0 to 5.0) works well on any NoobAI Checkpoint.

High weight or CFG may cause oversaturation unless using advanced tools.

DynamicThresholding (CFG-Fix) is effective to reduce oversaturation when using CFG 4.0 to 6.0.

DynamicThresholding settings: Mimic Scale 2.5 to 3.5, Mimic mode: Half Cosine Down, CFG Mode: Half Cosine Down.

Other tools to explore include Rescale CFG and Latent Modifier.

Try using Smooth Lighting Enhancer and Smooth Detailer Booster together for best results.

SMOOTH LIGHTING v.2!

Can be used on NoobAi/Illustrious/Pony Checkpoints.

NoobAI is proving to be very VERY susceptible to oversaturation and overexpose. This new version is an attempt to mitigate this a little bit while still giving improvement to lighting.

General use:

  • Weight between (-0.5 ~ 0.5) should be enough for 90% of anything you are trying to make!

  • Negative weight can be useful to mitigate oversaturation on some images.

  • Using Negative Weight (-1.0 ~ 0.0) will put an emphasis on add/strenghening shadows and dark colors while Positive Weight (0.0 ~ 1.0) will put an emphasis on add/strenghening light sources and bright colors.

  • For "Smooth Mix - NoobAI" use CFG between (2.0 ~ 4.0)! In fact, low CFG (3.0 ~ 5.0) works really well on any NoobAI Checkpoint!

  • High weight or high CFG with this Lora will easily lead to oversaturation/overexpose on NoobAI Checkpoints UNLESS you use advance tools, so only experience users should go for it.

Advanced use:

  • To reduce or eliminate oversaturation/overexpose the most effective way I have tested so far is using "DynamicThresholding (CFG-Fix)".

    • DynamicThresholding Settings: When Using CFG 4.0 ~ 6.0 - Mimic Scale: 2.5 ~ 3.5 / Mimic mode: Half Cosine Down / CFG Mode: Half Cosine Down

  • Every image on the showcase with this Lora weight 1.0 or CFG over 4.0 used DynamicThresholding.

  • Other tools include "Rescale CFG" and "Latent Modifier", but I haven't had enough time to test those. Feedback on those would be greatly appreciated.

With anyone has others suggestions on how to mitigate oversaturation be sure to share in the comments! =)

Have fun!

SMOOTH LIGHTING ENHANCER!

A companion to Smooth Detailer Booster with emphasis on lighting. It should work really well on Smooth Mix Checkpoint by giving more stuttle contrast between bright and dark areas while making light sources brighter.

Be sure to try Smooth Detailer Booster and Smooth Lighting Enhancer together!

Already tested on other Checkpoints with great results, so check them out! The weight can have a huge impact on the images depending on the Checkpoint used.

Have fun!

Poprzedni
Midjourney V6.1 meets FLUX 🖼️ [+ILLUSTRIOUS / SDXL] - ILLUSTRIOUS v0.5
Następny
smolMODE (Flux Pruned, Experimental) - FP16 (16.6Gb)

Szczegóły modelu

Typ modelu

LORA

Model bazowy

Illustrious

Wersja modelu

Smooth Lighting+

Hash modelu

f6fe5d752f

Dyskusja

Proszę się log in, aby dodać komentarz.

Obrazy autorstwa Smooth Lighting Enhancer (NoobAI/Illustrious) - Smooth Lighting+

Obrazy z illustrious

Obrazy z lighting

Obrazy z quality

Obrazy z narzędzie