๊ฒ€์€ ๋ฐฐ๊ฒฝ์— ๊ทน์ ์ธ ์กฐ๋ช… ์•„๋ž˜ ๊ธˆ๋ฐœ ๋จธ๋ฆฌ์™€ ํŒŒ๋ž€ ๋ˆˆ์„ ๊ฐ€์ง„ ๋งค์šฐ ์ƒ์„ธํ•œ ์—˜ํ”„ ๊ถ์ˆ˜.
๊ธฐ๊ณ„ ํŒ”๊ณผ ๋‚ ๊ฐœ ๋‹ฌ๋ฆฐ ํ—ฌ๋ฉง์„ ์ฐฉ์šฉํ•˜๊ณ  ๊ฒ€์„ ๋“  ๊ธˆ์ƒ‰ ๊ฐ‘์˜ท์˜ Malenia Blade of Miquella. ๋ถ‰์€ ๋จธ๋ฆฌ๊ฐ€ ์–ด๋‘์šด ๋ฐฐ๊ฒฝ์—์„œ ํ˜๋Ÿฌ๋‚ด๋ฆฌ๋ฉฐ ์œ ํ™” ์Šคํƒ€์ผ๋กœ ๋ Œ๋”๋ง๋จ.
๋ˆˆ ์œ„์— ๋””์ง€ํ„ธ ๊ธ€๋ฆฌ์น˜ ์˜ค๋ Œ์ง€ ๋ฐ”์ด์ €๋ฅผ ์“ด ๊ธˆ๋ฐœ ์†Œ๋…€, ํ‹ธ๊ณผ ์˜ค๋ Œ์ง€ ์ƒ‰์ƒ ํ…Œ๋งˆ ๋ฐ ํ™€๋กœ๊ทธ๋žจ ๊ทธ๋ž˜ํ”ฝ, ๋ƒ‰์ •ํ•œ ํ‘œ์ •์œผ๋กœ ์•‰์•„ ์žˆ๋Š” ์ž์„ธ

์ถ”์ฒœ ํ”„๋กฌํ”„ํŠธ

masterpiece,best quality,very aesthetic

1girl, solo, masterpiece, best quality, very aesthetic, looking at viewer, upper body

์ถ”์ฒœ ๋„ค๊ฑฐํ‹ฐ๋ธŒ ํ”„๋กฌํ”„ํŠธ

low quality,displeasing

low quality, displeasing, hair intakes, shiny hair

์ถ”์ฒœ ๋งค๊ฐœ๋ณ€์ˆ˜

samplers

Euler a

steps

25 - 28

cfg

3.5 - 4.5

ํŒ

์‚ฌ์‹ค์ ์ธ ํšจ๊ณผ๋ฅผ ์œ„ํ•ด์„œ๋Š” ๋†’์€ ๊ฐ€์ค‘์น˜์—์„œ โ€œrealisticโ€ ํƒœ๊ทธ๋ฅผ ์‚ฌ์šฉํ•˜๊ณ , โ€œphotorealisticโ€์€ ์งˆ๊ฐ ์กฐ์ •์šฉ์œผ๋กœ ๋‚ฎ์€ ๊ฐ€์ค‘์น˜์—์„œ ์‚ฌ์šฉํ•˜๋Š” ๊ฒƒ์„ ๊ถŒ์žฅํ•ฉ๋‹ˆ๋‹ค.

ppw ๊ณ ์ฐจ์› LoCon ๋ชจ๋ธ์€ ๋‹ค์–‘ํ•œ ๊ธฐ๋Šฅ LoRA ๋ฐ ์ฒดํฌํฌ์ธํŠธ์™€ ์œ ์—ฐํ•˜๊ฒŒ ์กฐํ•ฉํ•˜์—ฌ ํšจ๊ณผ ๊ฐ•๋„๋ฅผ ์ œ์–ดํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

์ด ๋ชจ๋ธ ์‹œ๋ฆฌ์ฆˆ๋Š” ๋ณด๋‹ค ์ ๊ณ  ๊ฐ„๋‹จํ•œ ํ’ˆ์งˆ ํ”„๋กฌํ”„ํŠธ๋กœ ์ด๋ฏธ์ง€ ํ’ˆ์งˆ์„ ํ–ฅ์ƒ์‹œ์ผœ ํ”„๋กฌํ”„ํŠธ์˜ ํ† ํฐ ๊ณต๊ฐ„์„ ์ ˆ์•ฝํ•ฉ๋‹ˆ๋‹ค.

๋Œ€๋ถ€๋ถ„์˜ ๋ชจ๋ธ ๋ฐ์ดํ„ฐ๋Š” AI ์ƒ์„ฑ ์ด๋ฏธ์ง€์ด๋ฉฐ 250์žฅ ์ด์ƒ์˜ ๊ณต๊ณต ๋ฏธ๋””์–ด์—์„œ ์ˆ˜๋™์œผ๋กœ ์„ ๋ณ„ํ•œ ์ด๋ฏธ์ง€๋„ ๊ฐœ๋… ๋ณด์™„์šฉ์œผ๋กœ ํฌํ•จ๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค.

์ด ๋ชจ๋ธ์€ ๋ฌด๋ฃŒ ์˜คํ”ˆ์†Œ์Šค์ด๋ฉฐ ๋™๋ฐ˜ ์ฒดํฌํฌ์ธํŠธ์™€ LoRA์˜ ๋ผ์ด์„ ์Šค๋ฅผ ์ค€์ˆ˜ํ•˜๋Š” ํ•œ ์ƒ์—…์  ์šฉ๋„๋กœ ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

ํ์‡„ ์†Œ์Šค ์ƒ์—…์  ์‚ฌ์šฉ, ๋ชจ๋ธ ํŒ๋งค ๋˜๋Š” ํ์‡„ ์†Œ์Šค ์ƒ์—… ๋ชจ๋ธ๊ณผ์˜ ํ•ฉ๋ณ‘์€ ๊ธˆ์ง€๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค.

v8

์งˆ๊ฐ ์—…๋ฐ์ดํŠธ๏ผš๋‹ค์Œ ํƒœ๊ทธ๋“ค์˜ ํ•™์Šต์„ ๊ฐ•ํ™”ํ–ˆ์Šต๋‹ˆ๋‹ค๏ผš

Texture Update: ๋‹ค์Œ ํƒœ๊ทธ๋“ค์ด ํ•™์Šต์—์„œ ๊ฐ•ํ™”๋˜์—ˆ์Šต๋‹ˆ๋‹ค:

realistic, photorealistic, flat color,
shiny skin, matte skin, shiny hair,

Danbooru ๋ฐ์ดํ„ฐ์…‹์—๋Š” โ€œ์‚ฌ์ง„โ€์ด๋‚˜ โ€œ์‚ฌ์ง„๊ณผ ์œ ์‚ฌํ•œ ์Šคํƒ€์ผโ€์„ ์„ค๋ช…ํ•˜๋Š” ์—ฌ๋Ÿฌ ํƒœ๊ทธ๊ฐ€ ์žˆ์Šต๋‹ˆ๋‹ค. ์ €๋Š” ํ•™์Šต ๋ฐ์ดํ„ฐ์…‹์—์„œ ์ด ๋ชจ๋“  ์ด๋ฏธ์ง€๋ฅผ "photorealistic"์œผ๋กœ ํ†ต์ผํ•˜์—ฌ ํƒœ๊น…ํ–ˆ์Šต๋‹ˆ๋‹ค. ํ•˜์ง€๋งŒ Danbooru ํ•™์Šต์…‹์œผ๋กœ ํ›ˆ๋ จ๋œ ๋Œ€๋ถ€๋ถ„์˜ SDXL ๋ชจ๋ธ์€ ์‚ฌ์‹ค์ ์ธ ์ด๋ฏธ์ง€๋ฅผ ์ž˜ ๊ทธ๋ฆฌ์ง€ ๋ชปํ•˜๋ฏ€๋กœ โ€œphotorealisticโ€์€ ๋‚ฎ์€ ๊ฐ€์ค‘์น˜์—์„œ๋งŒ ์งˆ๊ฐ ์กฐ์ •์„ ์œ„ํ•ด ์‚ฌ์šฉํ•˜๋Š” ๊ฒƒ์ด ๊ถŒ์žฅ๋ฉ๋‹ˆ๋‹ค. โ€œrealisticโ€์€ ๋†’์€ ๊ฐ€์ค‘์น˜์—์„œ๋„ ์ •์ƒ ์ž‘๋™ํ•ฉ๋‹ˆ๋‹ค.

Danbooru dataset contains multiple tags to describe "photo" or "photo-like styles". Iโ€™ve tagged all such images as โ€œphotorealisticโ€ in dataset.

However, most SDXL models trained on the Danbooru dataset do not render realistic images well. โ€œphotorealisticโ€ is only recommended at low weight, where it can help adjust texture rather than create realism images. The โ€œrealisticโ€ tag can work properly at higher weight.

๋น ๋ฅธ ์‹œ์ž‘ | Quick Start

์ด๊ฒƒ์€ ๋ฌด์—‡์ธ๊ฐ€? | What is this?

  • Pony: People's Works (ppw)๋Š” ์‹คํ—˜์ ์ธ ๋ฏธ์„ธ ์กฐ์ • ๋ชจ๋ธ ์‹œ๋ฆฌ์ฆˆ๋กœ, ๋ฐ์ดํ„ฐ์…‹์˜ ์•ฝ 85%๊ฐ€ CivitAI์—์„œ ์‚ฌ์šฉ์ž๊ฐ€ ๊ฒŒ์‹œํ•œ AI ์ƒ์„ฑ ์ด๋ฏธ์ง€๋กœ ์ˆ˜์ง‘๋˜์—ˆ์Šต๋‹ˆ๋‹ค. ์ดˆ๊ธฐ ppw ๋ฐ์ดํ„ฐ์…‹์€ pony v6๊ฐ€ ์ƒ์„ฑํ•œ ์ด๋ฏธ์ง€ ๊ธฐ๋ฐ˜์œผ๋กœ ๊ตฌ์ถ•๋˜์–ด, ์ด ์‹œ๋ฆฌ์ฆˆ ๋ชจ๋ธ์˜ ์ถœ๋ ฅ๋ฌผ ์—ญ์‹œ pony diffusion์˜ ํŠน์ง•์„ ๊ฐ–๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค.

  • ์ด ์‹œ๋ฆฌ์ฆˆ ๋ชจ๋ธ์€ ํ‘œ์ค€ Danbooru ํƒœ๊ทธ๋ฅผ ์‚ฌ์šฉํ•˜๋ฉฐ, ์ฃผ๋กœ ์ค‘๊ฐ„ ๋ฐ ๊ทผ์ ‘ ๊ฑฐ๋ฆฌ์˜ ์Šคํƒ€์ผํ™”๋œ ์ดˆ์ƒํ™” ์ƒ์„ฑ์„ ์ „๋ฌธ์œผ๋กœ ํ•ฉ๋‹ˆ๋‹ค. ์ฃผ๋œ ๊ธฐ๋Šฅ์€ ๊ธฐ๋ณธ ๋ชจ๋ธ์ด ์•„ํ‹ฐ์ŠคํŠธ ํ‚ค์›Œ๋“œ ์—†์ด, ์ ์€ ํ’ˆ์งˆ ํ”„๋กฌํ”„ํŠธ๋งŒ์œผ๋กœ ์ƒ๋Œ€์ ์œผ๋กœ ์•ˆ์ •์ ์ธ ์ด๋ฏธ์ง€ ํ’ˆ์งˆ์„ ์–ป๋„๋ก ํ•˜์—ฌ ํ”„๋กฌํ”„ํŠธ ํ† ํฐ ๊ณต๊ฐ„์„ ์ ˆ์•ฝํ•˜๋Š” ๊ฒƒ์ž…๋‹ˆ๋‹ค.
    ๋ณธ ๋ชจ๋ธ์€ ์Šคํƒ€์ผ LoRA๊ฐ€ ์•„๋‹ˆ๋ฉฐ, ๋‹ค์–‘ํ•œ ํ”„๋กฌํ”„ํŠธ์™€ ์ƒ์„ฑ ์กฐ๊ฑด์— ๋”ฐ๋ผ ๋ฏธ๋ฌ˜ํ•œ ์Šคํƒ€์ผ ์ฐจ์ด๊ฐ€ ์žˆ์„ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

  • Pony: People's Works (ppw)์€ ์•ฝ 85%์˜ ๋ฐ์ดํ„ฐ์…‹์ด CivitAI ์‚ฌ์šฉ์ž ๊ฒŒ์‹œ AI ์ด๋ฏธ์ง€๋กœ ๊ตฌ์„ฑ๋œ ์‹คํ—˜์  ๋ฏธ์„ธ ์กฐ์ • ๋ชจ๋ธ ์‹œ๋ฆฌ์ฆˆ์ž…๋‹ˆ๋‹ค. ์ดˆ๊ธฐ ppw ๋ฐ์ดํ„ฐ์…‹์€ Pony V6 ์ƒ์„ฑ ์ด๋ฏธ์ง€ ๊ธฐ๋ฐ˜์œผ๋กœ ๋งŒ๋“ค์–ด์ ธ Pony Diffusion ํŠน์„ฑ์„ ํฌํ•จํ•ฉ๋‹ˆ๋‹ค.

  • ์ด ์‹œ๋ฆฌ์ฆˆ๋Š” ํ‘œ์ค€ Danbooru ํƒœ๊ทธ๋ฅผ ์‚ฌ์šฉํ•˜๋ฉฐ ์ค‘๊ฐ„ ๋ฐ ๊ทผ๊ฑฐ๋ฆฌ ์Šคํƒ€์ผํ™” ์ดˆ์ƒํ™” ์ƒ์„ฑ์— ์ตœ์ ํ™”๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค. ์ด ๋ชจ๋ธ ์‹œ๋ฆฌ์ฆˆ์˜ ์ฃผ๋œ ๋ชฉ์ ์€ ์•„ํ‹ฐ์ŠคํŠธ ํ‚ค์›Œ๋“œ ์—†์ด ์•ˆ์ •์ ์ธ ์ด๋ฏธ์ง€ ํ’ˆ์งˆ์„ ์ œ๊ณตํ•˜๊ณ  ํ”„๋กฌํ”„ํŠธ ํ† ํฐ ๊ณต๊ฐ„์„ ์ ˆ์•ฝํ•˜๋Š” ๊ฒƒ์ž…๋‹ˆ๋‹ค.
    ์Šคํƒ€์ผ LoRA๊ฐ€ ์•„๋‹ˆ๋ฏ€๋กœ ํ”„๋กฌํ”„ํŠธ ๋ฐ ์กฐ๊ฑด์— ๋”ฐ๋ผ ๋ฏธ๋ฌ˜ํ•œ ์Šคํƒ€์ผ ์ฐจ์ด๊ฐ€ ์žˆ์Šต๋‹ˆ๋‹ค.

๋ฒ„์ „ ์ •๋ณด | Version Info.

  • ๋ณธ ํŽ˜์ด์ง€๋Š” ppw์˜ ๊ณ ์ฐจ์› LoCon ๋ฒ„์ „์„ ์ œ๊ณตํ•˜๋ฉฐ, ํ”„๋กœ์ ํŠธ์˜ ๋ฉ”์ธ ํŽ˜์ด์ง€ ์—ญํ• ์„ ํ•ฉ๋‹ˆ๋‹ค.

  • LoCon ๋ฒ„์ „ ppw๋Š” ๊ฐ์ข… ๊ธฐ๋Šฅ LoRA ๋ฐ ๋ฒ ์ด์Šค ๋ชจ๋ธ๊ณผ ์œ ์—ฐํ•˜๊ฒŒ ์กฐํ•ฉ ๊ฐ€๋Šฅํ•˜๋ฉฐ, ํšจ๊ณผ ๊ฐ•๋„ ์กฐ์ ˆ์ด ๋›ฐ์–ด๋‚ฉ๋‹ˆ๋‹ค. ๊ณ ์ฐจ์› ๋ฒ„์ „์€ ์ผ๋ฐ˜ํ™” ๋ฐ ์„ธ๋ถ€ ํ‘œํ˜„๋ ฅ์ด ๋” ๊ฐ•๋ ฅํ•˜์ง€๋งŒ ์ €์žฅ ๊ณต๊ฐ„๊ณผ ์—ฐ์‚ฐ ์ž์›์„ ๋” ๋งŽ์ด ์‚ฌ์šฉํ•ฉ๋‹ˆ๋‹ค.

  • ์ฃผ๋กœ ์˜จ๋ผ์ธ ์ƒ์„ฑ ์„œ๋น„์Šค ๋˜๋Š” ๊ณ ์„ฑ๋Šฅ PC๋ฅผ ๊ฐ€์ง„ ์‚ฌ์šฉ์ž์˜ ๋กœ์ปฌ ์ƒ์„ฑ์— ์ ํ•ฉํ•ฉ๋‹ˆ๋‹ค.

  • This page features the high dim LoCon version models of ppw, which also serves as the main page of this project.

  • The LoCon versions of ppw can be flexibly combined with various functional LoRAs and checkpoints, offering greater controllability over effect weight. The high dimension versions provide stronger generalization and more detailed rendering, but require more storage and computational resources.

  • They are mainly intended for online generation services and local use by users with high-performance PCs.


๊ฒฝ๋Ÿ‰ํ™” LoCon | Lightweight LoCon ver.

๋ฒ ์ด์Šค ๋ชจ๋ธ ๋ฒ„์ „ | Checkpoint versions (Illustrious)

๋ฒ ์ด์Šค ๋ชจ๋ธ ๋ฒ„์ „ | Checkpoint versions (NoobAI)

๋”๋ณด๊ธฐ | more

์‚ฌ์šฉ ๋ฐฉ๋ฒ• | Usage

๊ธ์ • ํ”„๋กฌํ”„ํŠธ:

masterpiece, best quality, very aesthetic

๋ถ€์ • ํ”„๋กฌํ”„ํŠธ:

low quality, displeasing

์—…๋ฐ์ดํŠธ ๊ธฐ๋ก | Change log

v7

v7 ๋ฒ„์ „์€ ๋ฐ์ดํ„ฐ์…‹ ๊ตฌ์กฐ์— ์ƒ๋‹นํ•œ ์กฐ์ •์„ ๊ฐ€ํ–ˆ์œผ๋ฉฐ, ๋‹ค๋ฅธ ํ•™์Šต ๋งค๊ฐœ๋ณ€์ˆ˜์™€ ์ „๋žต์„ ์‚ฌ์šฉํ•˜์˜€๊ธฐ ๋•Œ๋ฌธ์— v7์ด ์ด์ „ ๋ฒ„์ „๋ณด๋‹ค ์•ˆ์ •์ ์ด์ง€ ์•Š์„ ์ˆ˜๋„ ์žˆ์Šต๋‹ˆ๋‹ค.

The v7 version has undergone significant structural adjustments to the dataset, and utilizes different training parameters and strategies. As a result, v7 may be less stable than the previous versions.

v-pred ๋ชจ๋ธ์€ CivitAI ์˜จ๋ผ์ธ ์ƒ์„ฑ๊ธฐ์—์„œ์˜ ์„ฑ๋Šฅ์ด TensorArt ์˜จ๋ผ์ธ ์ƒ์„ฑ๊ธฐ์™€ ์™„์ „ํžˆ ๋‹ค๋ฅด๋ฉฐ, ๋™์ผํ•œ ํŒŒ๋ผ๋ฏธํ„ฐ๋กœ๋Š” ์žฌํ˜„์ด ๋ถˆ๊ฐ€๋Šฅํ•ฉ๋‹ˆ๋‹ค. ์ด์œ ๋Š” ์•Œ ์ˆ˜ ์—†์Šต๋‹ˆ๋‹ค...

The v-pred model's performance on the CivitAI online generator is completely different from online generation on TensorArt. The results are entirely unreproducible with a same parameters. I have no idea why...

TensorArt ๋ฒ„์ „ CivitAI ๋™์ผ ํŒŒ๋ผ๋ฏธํ„ฐ ๋ฒ„์ „ ๋ฐ ๋” ๋†’์€ ๊ฐ€์ค‘์น˜ ๋ฒ„์ „์˜ ๋น„๊ต ์ด๋ฏธ์ง€

v7 ๋ฒ„์ „ ๊ฐœ์š”๏ผš

์ด๋Š” ์ด์ „ ์ž‘์—… ๋ฐ์ดํ„ฐ์…‹์„ ๊ธฐ๋ฐ˜์œผ๋กœ ๋ฐœ์ „๋œ ์ด๋ฏธ์ง€ ํ’ˆ์งˆ LoCon์œผ๋กœ, ์•ฝ 90%-95% ๋ฐ์ดํ„ฐ๊ฐ€ CivitAI์—์„œ ๊ฒŒ์‹œ๋œ ์ด๋ฏธ์ง€์ž…๋‹ˆ๋‹ค.

์ด ๋ชจ๋ธ์€ ์•„ํ‹ฐ์ŠคํŠธ ํ‚ค์›Œ๋“œ ์—†์ด, ์ ์€ ํ’ˆ์งˆ ํ”„๋กฌํ”„ํŠธ๋กœ ์ƒ๋Œ€์ ์œผ๋กœ ์•ˆ์ •์ ์ธ ์ด๋ฏธ์ง€ ํ’ˆ์งˆ์„ ์ œ๊ณตํ•˜๋ฉฐ, ๋” ๋งŽ์€ ํ† ํฐ ๊ณต๊ฐ„์„ ์ ˆ์•ฝํ•ฉ๋‹ˆ๋‹ค. ๋˜ํ•œ ์ผ๋ถ€ ๋ชจ๋ธ ๊ณ ์œ ์˜ ์ƒ์„ฑ ๊ฒฐํ•จ์„ ๊ณ ์น  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.(๋‹จ, ์† ์ œ์™ธ)

๋ฐ์ดํ„ฐ์…‹ ์„ ํƒ์œผ๋กœ ์ธํ•ด ์ƒ์„ฑ๋œ ์ด๋ฏธ์ง€๋Š” Pony ์Šคํƒ€์ผ ํŠน์„ฑ์„ ๊ฐ–์Šต๋‹ˆ๋‹ค. ๊ทธ๋Ÿฌ๋‚˜ ํŠน์ • ์•„ํ‹ฐ์ŠคํŠธ, ์Šคํƒ€์ผ ๋˜๋Š” ํ™”ํ’์„ ์ง€ํ–ฅํ•˜์ง€ ์•Š์•„ ํ”„๋กฌํ”„ํŠธ ๋ฐ ์กฐ๊ฑด์— ๋”ฐ๋ผ ๋ฏธ๋ฌ˜ํ•œ ์Šคํƒ€์ผ ์ฐจ์ด๊ฐ€ ์žˆ์„ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

This is a generation quality LoCon developed based on the dataset from the previous work. About 90%-95% of the image data comes from CivitAI.

It allows models to achieve relatively stable image quality without artist tags or using long quality prompts, freeing up more token space. Additionally, it can fix some inherent generation flaws of the model. (except for hands)

Due to the dataset selection, the generated images exhibit a Pony-like style. However, since it does not reference any specific artist, style, or painting technique, there may be subtle stylistic variations depending on different prompts and checkpoint conditions.

๋ฐ์ดํ„ฐ์…‹ ์ถœ์ฒ˜ ๋ฐ ๋ผ์ด์„ ์Šค | Dataset Source & License

  • ๋ฐ์ดํ„ฐ์…‹ ๋‚ด ๋ชจ๋“  ์ด๋ฏธ์ง€๋Š” ์ €์ž๊ฐ€ ์ง์ ‘ ์ˆ˜๋™ ์„ ๋ณ„, ๋ถ„๋ฅ˜, ์ฃผ์„ ํŽธ์ง‘ํ–ˆ์œผ๋ฉฐ ์ˆ˜๋ฐฑ ์žฅ์€ ์ˆ˜๋™์œผ๋กœ ํŽธ์ง‘ ๋ฐ ์ˆ˜์ •๋˜์—ˆ์Šต๋‹ˆ๋‹ค.

  • ์ด ๋ชจ๋ธ์€ ๋ฌด๋ฃŒ ์˜คํ”ˆ์†Œ์Šค ๋ชจ๋ธ๋กœ, ์‚ฌ์šฉ์ž๋Š” ๊ฐœ์ธ ์žฅ๋น„์— ์ž์œ ๋กญ๊ฒŒ ๋ฐฐํฌํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ์ €์ž๋Š” ๋ชจ๋ธ ํŒ๋งค๋กœ๋ถ€ํ„ฐ ์–ด๋– ํ•œ ๋ณด์ƒ๋„ ๋ฐ›์ง€ ์•Š์Šต๋‹ˆ๋‹ค. ์ €์ž๋Š” ๋ณธ ๋ชจ๋ธ์„ ์ƒ์—… ์ƒ์„ฑ ์„œ๋น„์Šค ํ˜น์€ ์ƒ์—…์  ์ด๋ฏธ์ง€ ์ƒ์„ฑ์— ์‚ฌ์šฉํ•˜๋Š” ๊ฒƒ์„ ์ œํ•œํ•˜์ง€ ์•Š์œผ๋‚˜ ์‚ฌ์šฉ๋˜๋Š” ์ฒดํฌํฌ์ธํŠธ ๋ฐ ๋‹ค๋ฅธ LoRA์˜ ๋ผ์ด์„ ์Šค ์ œํ•œ ์‚ฌํ•ญ์„ ๋ฐ˜๋“œ์‹œ ํ™•์ธํ•ด ์ฃผ์‹ญ์‹œ์˜ค.

  • ๋ฐ์ดํ„ฐ์…‹์˜ ์•ฝ 90%-95%๋Š” AI ์ƒ์„ฑ ์ด๋ฏธ์ง€์ด๋‚˜ ์•ฝ 250์žฅ ์ด์ƒ์˜ ๊ณต๊ฐœ ๋ฏธ๋””์–ด, ๋‰ด์Šค ๋ฏธ๋””์–ด, ์ถœํŒ๋ฌผ์—์„œ ์ˆ˜์ง‘ํ•œ ์ด๋ฏธ์ง€๊ฐ€ ๊ฐœ๋… ๋ณด์™„์„ ์œ„ํ•ด ํฌํ•จ๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค. ํ–ฅํ›„ ๋ฒ„์ „์—์„œ ์ ์ฐจ ๊ต์ฒด๋  ์˜ˆ์ •์ด๋‹ˆ ์ƒ์—…์šฉ ์‚ฌ์šฉ์ž๋Š” ๊ด€๋ จ ์œ„ํ—˜์„ ์Šค์Šค๋กœ ์ฃผ์˜ํ•ด ์ฃผ์‹ญ์‹œ์˜ค.

    ๋ณธ ๋ฐ์ดํ„ฐ์…‹์€ ํŠน์ • ๊ฐœ์ธ ์ž‘๊ฐ€ ๋ฐ์ดํ„ฐ๋ฅผ ํฌํ•จํ•˜์ง€ ์•Š์œผ๋ฉฐ, ์–ด๋–ค ์ž‘๊ฐ€ ์ •๋ณด๋„ ํ‘œ์‹œํ•˜์ง€ ์•Š์•˜์Šต๋‹ˆ๋‹ค(๋‹จ AI์˜ ์˜คํƒœ๊ทธ ๊ฐ€๋Šฅ์„ฑ์€ ์žˆ์Šต๋‹ˆ๋‹ค).

  • ๋˜ํ•œ, ๋ณธ ๋ชจ๋ธ์€ ํ์‡„ ์†Œ์Šค ์ƒ์—…์šฉ, ๋ชจ๋ธ ํŒ๋งค, ํ์‡„ ์†Œ์Šค ์ƒ์—… ๋ชจ๋ธ ํ•ฉ๋ณ‘์— ์‚ฌ์šฉ์ด ๊ธˆ์ง€๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค. ์˜คํ”ˆ์†Œ์Šค ํ•ฉ๋ณ‘ ๋ชจ๋ธ๋กœ ์ด๋ฏธ์ง€ ์ƒ์„ฑ ์„œ๋น„์Šค๋ฅผ ์ œ๊ณตํ•˜๋Š” ๊ฒฝ์šฐ์—๋Š” ์ œํ•œ์ด ์—†์ง€๋งŒ, ํ•ฉ๋ณ‘ ๋ชจ๋ธ ์ถœ์ฒ˜ ํ‘œ๊ธฐ๋ฅผ ๊ถŒ์žฅํ•ฉ๋‹ˆ๋‹ค.

  • Every image in the dataset has been manually selected, categorized, and annotated by the author. Additionally, hundreds of the images have been manually edited and corrected.

  • This model is free and open-source model, allowing users to deploy it on their personal devices. The author does not receive any compensation from selling the model. The author does not impose restrictions on using this model for commercial image generation services or generating images for commercial purposes. However, please be mindful of the license restrictions of the Checkpoint and other LoRAs used alongside this model.

  • Approximately 90%-95% of the dataset consists of AI-generated images. However, around 250+ images have been collected from public media, news outlets, and publications to supplement concepts. Future versions will gradually replace these materials. Users with commercial intentions should be aware of the potential risks.

    This dataset does not include training data from any individual artist, nor does it contain explicit artist attributions (though AI mistagging cannot be entirely ruled out).

  • Additionally, this model is not permitted for use in closed-source commercial applications, model resales, or merged into closed-source commercial models. There are no restrictions on open-source merged models being used for image generation services, but it is recommended to credit the sources of any merged models.

์ด์ „
Holy Mix [illustriousXL] ๊ณ ๋Œ€๋น„ ์• ๋‹ˆ๋ฉ”์ด์…˜ ์ฒดํฌํฌ์ธํŠธ - v1
๋‹ค์Œ
์ดˆํ˜„์‹ค ์• ๋‹ˆ๋ฉ” - 3k ์Šคํ…

๋ชจ๋ธ ์„ธ๋ถ€์‚ฌํ•ญ

๋ชจ๋ธ ์œ ํ˜•

LORA

๊ธฐ๋ณธ ๋ชจ๋ธ

NoobAI

๋ชจ๋ธ ๋ฒ„์ „

v7_noobVv1.0

๋ชจ๋ธ ํ•ด์‹œ

07ab55e611

์ œ์ž‘์ž

ํ† ๋ก 

๋Œ“๊ธ€์„ ๋‚จ๊ธฐ๋ ค๋ฉด log inํ•˜์„ธ์š”.