If you like, pleasehttps://ko-fi.com/bolero537This model was created by merging the following models:
V1.2 ①＋②＋③
V1.3 V1.2+④
V1.4 V1.3+⑤
V1.5 V1.4＋Lora(straight hair + eyelashes + lkp)
V1.51 V1.5 + Lora
V1.70 V1.51 + ⑥ + ⑦ + lora
私が作成したloraとの相性はV1.4の方がいいかもしれません。
V1.51は指の破綻軽減を目指したものですが、あまり変わらないかも。
① pony_pencil-XL - v1.0.2 | Stable Diffusion Checkpoint | Civitai
② MomoiroPony - v1.4 | Stable Diffusion Checkpoint | Civitai
③ Mala Anime Mix NSFW PonyXL - v2.0 | Stable Diffusion Checkpoint | Civitai
④ ChacolEbaraMixXL - v2.1 | Stable Diffusion Checkpoint | Civitai
⑤ EvaClausMix Pony XL - v1.0 | Stable Diffusion Checkpoint | Civitai
⑥ WAI-ANI-NSFW-PONYXL - v11 | Stable Diffusion Checkpoint | Civitai
⑦ Kohaku-XL Delta - rev1 | Stable Diffusion XL Checkpoint | Civitai
Description
FAQ
Comments (18)
This model is delicate and compatible with Lora. It's a very cool model!
2024年6月8日~我用過站上大部分的Pony分支大模型，這個是我用過最好的。我用它完成了一本195頁的同人漫畫。名字就叫"廬山昇龍霸強姦雅典娜"。
以下為漫畫連結~ 感謝作者的大模型和Lora。
Thank you so much~
こんにちは。このモデル気に入ってLora作りたいのですがノイズのような画像になります。他のponyモデルで全く同じ設定でLoraを作ると、正常になるのですが。。。 VAEは試しました。正体不明すぎて途方に暮れているので良ければこのモデルでLoraを作る時のJsonを公開していただけないでしょうか？
私の作っているLoraは別のモデルを使用しているので、原因はわからないです。ちょっと試してみます。
問題なくできました。Loraと一緒にアップロードしておきます。日焼け跡のLoraです。
@unknown_bolero537 早い！ありがとうございますjsonを見るとコサインが原因かな。。。それともNoise offset設定してるからかも
⑤のEvaClausMix Pony XL の注意書きにLoraできないかもって書いてあるからそれも考えられます。だからboleroのバージョン1.3でも作ってみます。ありがとう！！
@Unkoj その後はどうですか。同じ現象かわかりませんが、色が抜けたような画像になるロラができました。正則化画像を設定すると発生します。
If updating the v1.5 model, it is recommended to re-integrate it.
Really wanted to try this but the color gets washed out on fooocus even though I have tried various sample settings :(
same
That's because the model doesn't have a VAE baked in, which is very unusual. I'm not sure why the model creator did this, because it makes the model really annoying to use.
Without a VAE, an image will look nice in the previews, but then suddenly turns grey/brown and lifeless as soon as it's done. Most models have a VAE built in, so it's almost never an issue, but when they don't, you need to set your own VAE in order to use it. (You'll have to download it. Search "SDXL vae low memory" or something like that. Pony is technically a SDXL model.) If fooocus doesn't let you pick a custom VAE, you won't be able to use this model at all. 😐
Honestly, it's enough of a hassle to have to manually set a VAE (I always forget until I generate a dead image) and then also unset it when I switch to anything else (because it can create other problems if you leave it on, especially if you switch to a SD1.5 model) that I usually don't even bother with models that don't include them, even if I think they're great otherwise.
So, I think it might be worth skipping this checkpoint.
which VAE do you use?
I'm using sdxl_vae.
This is a great model - thank you!
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.