catboxanon

NAI Tags+

This userscript injects artists and tags that do not conform to "AI alignment" standards into NovelAI's suggested imagegen tags.

Prerequisites

You will need a userscript extension. Below is a suggestion.

• https://violentmonkey.github.io

Installation

catboxanon

// ==UserScript==
// @name        4chan pixeldrain URL parser
// @namespace   4chanpixeldrainparse
// @match       https://boards.4chan.org/*/thread/*
// @grant       none
// @version     1.0.3
// @author      Anonymous
// @description Parse pixeldrain URLs on 4chan
// @updateURL   https://gist.github.com/raw/6fedeaacb169bc89bc5186bd00741b4a/4chan-pixeldrain.user.js
// @downloadURL https://gist.github.com/raw/6fedeaacb169bc89bc5186bd00741b4a/4chan-pixeldrain.user.js

catboxanon

NOTICE

I no longer plan to maintain this userscript, since as of 2025-03-01, NovelAI has added a native shortcut for attention emphasis. It only took them 1 year and 3 months!

The shortcut is triggered by highlighting anything in the prompt, and pressing [ to de-emphasize, or Shift + [ to emphasize.

Auto-saving images has also been a native feature for quite a while now, accessible by clicking on the goose icon in the top left.

Of the other two features exclusive to this script, generating forever and wildcards, the former still works, while the latter is broken.

catboxanon

// ==UserScript==
// @name        DALL-3 Save
// @namespace   bingdall3save
// @match       https://www.bing.com/images/create*
// @version     1.0.2
// @author      Anonymous
// @updateURL   https://gist.github.com/raw/fa297d9a104b7dc63f92b444cbe51b39/dall-3-save.user.js
// @downloadURL https://gist.github.com/raw/fa297d9a104b7dc63f92b444cbe51b39/dall-3-save.user.js
// @require     https://cdn.jsdelivr.net/npm/file-saver
// ==/UserScript==

catboxanon

import torch

from modules import script_callbacks, shared

def on_model_loaded(sd_model):
    if hasattr(shared.opts, 'clip_tensor_fix_enabled') and shared.opts.data.get('clip_tensor_fix_enabled', False): # type: ignore
        try:
            t1 = None
            t2 = torch.arange(0, 77, dtype=torch.int64).unsqueeze(0)

catboxanon

import gradio as gr
import numpy as np
import torch
from PIL import Image
from scipy.ndimage import gaussian_filter
from skimage.transform import resize

import modules.scripts as scripts
from modules import devices, script_callbacks, shared
from modules.processing import StableDiffusionProcessingTxt2Img

catboxanon

import argparse
from glob import glob
from pathlib import Path

import numpy as np
from tqdm import tqdm


def read_metadata_from_safetensors(filename):
    import json

catboxanon

from pathlib import Path
from tqdm import tqdm
import argparse

def read_metadata_from_safetensors(filename):
	import json

	with open(filename, mode="rb") as file:
		metadata_len = file.read(8)
		metadata_len = int.from_bytes(metadata_len, "little")

catboxanon

import os
from typing import Callable, List, Union

import gradio as gr

from modules import scripts
from modules.processing import StableDiffusionProcessing, StableDiffusionProcessingTxt2Img
from modules.script_callbacks import (CFGDenoisedParams, CFGDenoiserParams,
									  on_cfg_denoised, on_cfg_denoiser)

catboxanon

import os
import torch

import gradio as gr

from modules import script_callbacks, scripts, shared


class CLIPBlend:
    def __init__(self):