模型/badpicture negative embedding for ChilloutMix - 75 Vector Version

badpicture negative embedding for ChilloutMix - 75 Vector Version

|
7/2/2024
|
12:25:12 AM
| Discussion

推薦提示詞

8k,masterpiece,best quality

推薦反向提示詞

bad-picture-chill-75v,bad-picture-chill-32v,bad-picture-chill-1v

推薦參數

samplers

Euler, DPM++ 2M Karras, Euler a, DPM++ SDE Karras, DPM++ SDE

steps

20 - 35

cfg

5 - 7

other models

japaneseDollLikeness_v10, koreanDollLikenesss_v10

提示

Download the file and place it in the embeddings folder for A1111 Webui.

Use the embedding by entering its name in the 'negative prompt' box.

For higher detail, use the 75 vector version but be cautious of 'deep-frying' effects.

版本亮點

Trained until step 2101 @ 4batch 5grad with a variable learning rate using an extension for WebUI that has advanced TI training features. The training .zip contains saves with lesser and higher step counts.

創作者贊助

March 17, 2023 edit: quick note on how to use a negative embeddings. You download the file and put it into your embeddings folder. Since I use A1111 Webui, that is stable-diffusion-webui/embeddings. If you use a different program you will have to check with its documentation on where to put embeddings.

After you have put the file in the appropriate folder, you then use the embedding by entering its name into the "negative prompt" box instead of the normal prompt. The name, by default, will be either bad-picture-chill-75v, bad-picture-chill-32v, or bad-picture-chill-1v, depending on which one you downloaded. If you rename the file, that will change the activation word to whatever you renamed the file. So if you renamed "bad-picture-chill-75v.pt" to "pizza-is-great.pt" the activation phrase changes to "pizza-is-great"

End of edit.

This is a "bad prompt" style embedding- it is trained to be used in the negative prompt. It is made specifically for ChilloutMix (version chilloutmix_NiPrunedFp32Fix.safetensors [fc2511737a]), so it may or may not work well on a non-chilloutmix model. The prompts used for the pictures were mostly taken from the comment section on the ChilloutMix page.

It comes in 1 vector, 32 vector, and 75 vector versions- the higher you go, the more detail you will be able to squeeze out, but the closer you will get to deep-frying your image. Check out the comparison images that I uploaded as part of the 75 vector version- you'll probably have to right-click and "open image in new tab" to properly zoom in.

The 'Training Images" download for each version is a .zip that has all the saved .pt versions that were generated during the training. Look there if you want a version that has more or less steps. If you want the actual training images, click on the "Actual Training Images" version which should be below the 1V version of the embedding.

Preview images: all preview images were generated at 512x512 with highres enabled @ a factor of 2 and 0.75 denoising, resulting in a final image of 1024x1024. Model: chilloutmix_NiPrunedFp32Fix.safetensors [fc2511737a]

Sometimes the original prompt or original negative prompt will include a LORA or embedding- here are links to them. Note: these are not used in the single image previews, only the original prompts that I borrowed and put into the stacked comparison images.

Japanese Dolllikeness lora (robot lady image)

easynegative embed (armored lady) (impossibly tight black bodysuit lady)

bad-prompt embed (evil dogzilla looking thing)

ng_deepnegative_v1_75 embed (fox ears lady)

上一個
AbsoluteReality - v181
下一個
AnythingQingMix - v30

模型詳情

模型類型

TextualInversion

基礎模型

SD 1.5

模型版本

75 Vector Version

模型雜湊值

7d9cc5f549

訓練詞彙

bad-picture-chill-75v

討論

log in以發表評論。