Sign In

Old Newspaper Style

874
4.6k
6.2k
73
Verified:
SafeTensor
Type
LoRA
Stats
4,571
6,157
Reviews
Published
Apr 25, 2023
Base Model
SD 1.5
Trigger Words
npzw
npzw, (text), monochrome
npzw, (text), sepia
Hash
AutoV2
18AFF08228
default creator card background decoration
Maintenance Mode Badge May 2023
gahara42's Avatar
gahara42

This is a style LoRA that will give your images a old newspaper front page look. Header on top, text around the frame, where your object/subject will be inside the frame or with text accurately around your object/subject.

I want to thank @Hollowstrawberry for the collab he made. And also @rockerBOO, @ChameleonAI, @zykurv, @IndolentCat for their tips and tricks! Please check their creations as well!

The reviews and feedback will be very valuable for me.

Trigger word:

The main trigger word for style is npzw. You will additionally need to add to prompt the next words: (text), monochrome, or sepia. Keep in mind that depending on the model you using you may need to adjust weights of those words. Most often for anime models you want like: (npzw:1.1), (text:1.1) at least. Adjust that depending on your results.

And most important remove text, monochrome, sepia, greyscale from your negative prompt. Because a lot of people using those words in negative prompt.

  • Avoid using colors other than black or white in prompts.

Red eyes will give your image a hue that will ruin the purpose of this style in my opinion. However feel free to do so if you want. Also if 1girl / 1boy doesn't work for your character LoRA try using woman / man instead.

Use additional modifiers to frame the body of person: full body, upper body, cowboy shot, portrait

Weights:

This LoRA was trained on SD1.5. So it works with weight :1 absolutely fine. Adjust it if needed when combining with other LoRAs.

I'm not sure what name LoRA will get on downloading, so please make sure you using the proper name in prompt of the file that you downloaded if you going to copy prompts from previews.

Negative embeddings:

I think i used only one negative embedding during testing: