私のお気に入りの2つのモデルをマージした物です。
This is a merging of two of my favorite models.
*********************************************
マージ元 Merge from
Colorburst XL pony1.0
@agress74
https://civarchive.com/models/341261?modelVersionId=384017
Async's MIX XL
@asyncforai414
https://civarchive.com/models/266052?modelVersionId=358988
*********************************************
Euler a ステップ30
高解像度補助:R-ESRGAN 4x+ Anime6B,倍率1.5倍、ステップ15、デノイジングストレングス0.5
ADetailerを使用しface_yolov8nで顔修正を行っています。
Euler a Step 30
High resolution auxiliary: R-ESRGAN 4x+ Anime6B, 1.5x magnification, step 15,Denoising strength 0.5
I am using ADtailer to modify the face with face_yolov8n.
*********************************************
"ADetailer"
(インストールは自己責任でお願いします)(Please install at your own risk.)
顔の詳細を修正する追加機能 Additional features to correct facial details
https://github.com/Bing-su/adetailer
*********************************************
推奨追加"LORA" Recommended addition "LORA"
<lora:bold-16oa:-0.5>,
sdxl-boldline.safetensors
@2vXpSwA7
https://huggingface.co/2vXpSwA7/iroiro-lora/tree/main/sdxl
"XL Outline / 轮廓线 pony/sdxl/kohaku"
@kitarz
https://civarchive.com/models/169170/xl-outline-ponysdxlkohaku
"DetailedEyes_XL"
@bdsqlsz
https://civarchive.com/models/120723/detailedeyesxl
*******************************
Embeddings
negativeXL_D
@rqdwdw
https://civarchive.com/models/118418?modelVersionId=134583
*******************************
"sd-forge-couple"
(インストールは自己責任でお願いします)(Please install at your own risk.)
エンターもしくは指定の文字でタイル処理できるforge専用の追加機能。
An additional feature exclusive to forge that allows you to process tiles with enter or specified characters.
Description
KntanMIX XLpony v1.4 baked VAE
私のお気に入りの2つのモデルをマージした物です。
This is a merging of two of my favorite models.
*********************************************
SDXL VAEが焼き付けされています、多少の描画の向上が有りますが、使用メモリの増大、描画速度の低下の可能性が有ります。
気になる方は、no VAEバージョンを使用してFP16VAEを追加してみて下さい
SDXL VAE is baked in, which improves rendering slightly, but may increase memory usage and slow down rendering speed.
If you are concerned, try using the no VAE version and adding FP16VAE.
*********************************************
マージ元 Merge from
Colorburst XL pony1.0
@agress74
https://civitai.com/models/341261?modelVersionId=384017
Async's MIX XL
@asyncforai414
https://civitai.com/models/266052?modelVersionId=358988
*********************************************
Euler a ステップ30
高解像度補助:R-ESRGAN 4x+ Anime6B,倍率1.5倍、ステップ15、デノイジングストレングス0.5
ADetailerを使用しface_yolov8nで顔修正を行っています。
Euler a Step 30
High resolution auxiliary: R-ESRGAN 4x+ Anime6B, 1.5x magnification, step 15,Denoising strength 0.5
I am using ADtailer to modify the face with face_yolov8n.
*********************************************
"ADetailer"
(インストールは自己責任でお願いします)(Please install at your own risk.)
顔の詳細を修正する追加機能 Additional features to correct facial details
https://github.com/Bing-su/adetailer
*********************************************
推奨追加"LORA" Recommended addition "LORA"
<lora:bold-16oa:-0.5>,
sdxl-boldline.safetensors
@2vXpSwA7
https://huggingface.co/2vXpSwA7/iroiro-lora/tree/main/sdxl
"XL Outline / 轮廓线 pony/sdxl/kohaku"
@kitarz
https://civitai.com/models/169170/xl-outline-ponysdxlkohaku
"DetailedEyes_XL"
@bdsqlsz
https://civitai.com/models/120723/detailedeyesxl
*******************************
Embeddings
negativeXL_D
@rqdwdw
https://civitai.com/models/118418?modelVersionId=134583
*******************************
"sd-forge-couple"
(インストールは自己責任でお願いします)(Please install at your own risk.)
エンターもしくは指定の文字でタイル処理できるforge専用の追加機能。
An additional feature exclusive to forge that allows you to process tiles with enter or specified characters.
FAQ
Comments (24)
Can we make this basemodel available onsite?
please remember me
Online generation is permitted but not adopted by CVITAI
@jibunsagasinotabi
Can it be changed to be available on civitai?
@sesaryo583senu
I give permission.
CVITAI has not adopted it.
Great checkpoint, thanks!
素晴らしいチェックポイントだ!
Can you trying merge a 8-step turbo version?
We have released the 1.4 turbo version. The drawing speed has improved by about 30%, but the drawing content seems to have deteriorated considerably.
@jibunsagasinotabi Are you using cfg 2 in 1.4 turbo version?
any model can be turned to a 2 step model at this point,
@glitter_fart The main trend is 8 steps, which is troublesome to remember every different base model, so I completely ignore the 4-step or 2-step basic model. But steps 2 and 4 are suitable for bitch size 4 to generate.
@jibunsagasinotabi Is you not merged into the baked version (i will avoid it)? Or merge lora directly?
@jibunsagasinotabi I've always wanted to test it with your model, but only if a turbo version can be provided...(I'm always embarrassed to merge other people's models, It would be more embarrassing if you merged yours😂)
@jibunsagasinotabi If merge turbo lora, recommended to include dpo.
@jibunsagasinotabi Thanks. I saw the turbo version... I'll try it first.🤔
In my environment, CFG4 step 8 was required.
@jibunsagasinotabi I feel that merge ratio of the turbo is not enough, and the light source has a halo effect? soft focus lens? I usually use a ratio of 1:4 (cfg and step) in turbo, lcm is 1:2. You can see four example pictures. Currently it seems that cfg 5 step 20 is normal...Try merge turbo lora up to 100%? I tested two versions, and felt like there was no turbo at all, they are the same picture. 😂
@ghostpaint
I'm testing it too, but it's not working at all, it seems to be failing.
@jibunsagasinotabi Hmm. wish you success. I like it, your japanese anime style (I’m not used to other styles pony base model)...😂
We have released the LCM Turbo version, which is a modified version of the Turbo version. It is still in testing, but it works with CFG1-4, steps 2-8, and the drawing speed has improved by 30-50%. The drawing quality is quite low.
@jibunsagasinotabi There are currently two methods. One is to merge the turbo version of the original author of pony, provided that you still remember the recipe. Another method is to merge turbo lora, but I don't know the actual operation. Maybe the problem is the latter. The quality is low. I don’t know what the reason is. I haven’t read any related articles. 😅
@jibunsagasinotabi I looked at the Lcm +turbo version you drew, and the blur is not as serious as before... Testing requires efficiency, especially XYZ...Maybe my requirements for image quality are not too high, so I don't feel very bad...👍
@ghostpaint
I'm currently testing it, but it seems that the CFG needs to be set quite low, around 1.1 to 1.4.
The steps should be around 4 to 6 or 8.
@jibunsagasinotabi CFG 1.1~1.4 can modify some sharp line sparts, with a retro animation tone, which should be the style you are used to. CFG 2 is to increase the sense of lines, I usually choose this. Step 8 can retain enough background details, and step 4 is usually used for realism, because most of realism is depth of field. CFG 3~4 is suitable for sketch and hand-drawn style.
@jibunsagasinotabi Some people will use detail lora to improve the lack of detail at low steps, but for me step 8 is enough.
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.
















