Sign In

Varos - Midas Gold Touch 👑

13
119
3.6k
8
Verified:
SafeTensor
Type
LoRA
Stats
119
3,597
1.7k
Reviews
Published
Aug 9, 2024
Base Model
Pony
Training
Steps: 9,980
Epochs: 20
Usage Tips
Clip Skip: 2
Trigger Words
60ld
Hash
AutoV2
C1E2CE0E26
Donator Badge
_Varos_'s Avatar
_Varos_

Turn everything into GOLD with this Morph LoRA


Use your creativity and let the Artist out of you. 🎨⠀


- How to use 🔑

Lora is best used with a Strength of 0.5 to 1.0

Trigger:

60ld (Main Triggerword for Everything)

"......." is made entirely from Gold (optional - can help with the creation)

Additionally trained with over 400 high-resolution real photos of women with golden body painting in all possible positions.

Gold Bodypaint (for pictures with golden body color)

very simple example of bodypainting Images and the possibility to remove some of the unwanted elements of the training images.

1024 x 1280, Seed: 885074228, Steps: 50, CFG: 10, Checkpoint: authumanPony4_v25

LoRA Strength 1.0, Sampler: dpmpp_sde - sgm_uniform

Raw Output, no Upscale...

The LoRA tends to overfit the face or eye area. However, this depends heavily on the checkpoint and the set strength of the LoRA. This can be avoided by using detailers or upscaling, better prompting or a different checkpoint. I have tested many different epochs intensively.

an example of a usable bodypaint creation

Positive:⠀

score_9, score_8_up, score_7_up,

60ld, sexy milf, long hair, smile, high heels, gold bodypaint, shiny, makeup, photo shooting, detailed face, detailed eyes, shiny skin, glossy skin, clean face, lipstick,

(bikini:1.1), sfw,

sunlight, outdoor, clean wooden floor, plants, flowers, pool,

Negative:

score_3_up, score_4_up, score_5_up,

dirt, nsfw,

1024 x 1280, Seed: 829886009, Steps: 50, CFG: 10, Checkpoint: authumanPony4_v25

LoRA Strength 0.8, Sampler: dpmpp_sde - sgm_uniform

Raw Output, no Upscale...

Upscale:⠀

Reduce the LoRA strength slightly when upscaling. This gives you better results

-

100% Bio 💚

My example Images that I create for the LoRAs are always 100% Native! That means:

⋆ without additional LoRAs

⋆ without detailer

⋆ without In/Out Painting

⋆ without ControlNet

Only upscaling and color correction is used

so you can best see what you can easily create yourself!

-

LoRA Training

660 very well cherry picked high-resolution images and tagged by hand!

The auto tagging unfortunately very often adds things that are not there at all

and thus worsens the final result

For Example What the Automatic Tagged:

Coffee Machine = weapon, gun, military machine, gatling gun,

Bridge = tree, scenery,

Man = no human, food,

Rose = blood,

Pony = pokemon creature

Hamster = animal, whiskers, cat,

1girl = male, male focus,

Airplane = fish,

Train = tank,

Training in 9980 steps with 20 Epochs.

Which unfortunately makes the creation expensive but the main thing is that

the result is good.

Of course, like many here, I could just make a quick LoRA with 500 Buzz, but quality is simply more important to me.

For you too?

💖 Made with love 💋

-

History / Definition / Inspiration

interesting links to further information⠀

https://en.wikipedia.org/wiki/Midas#Golden_Touch

Can someone really suffocate from being covered with gold paint?

-

-

If you like my work?

Please support Me on Ko-fi or Patreon 🎁