foreword
This article mainly explains the usage of Stable Diffusion (hereinafter referred to as SD) prompt words to help you generate higher quality pictures
This chapter mainly explains Vincent diagrams, other types of readers can explore by themselves. At the same time, this article mainly generates pictures in the form of Stable Diffusion Discard
If you have requirements for image privacy, image copyright issues, or need more powerful and more detailed image generation, you can directly deploy the Stable Diffusion Web UI website on your own server. The deployment method can directly search bloggers for related articles
Preparation
Join Stable Foundatin: Discord
After joining, we can click these bots to enter the group, then click the dialog box at the bottom and enter /dream
, and click Enter, and then we can enter the prompt words to generate a picture
Then we click here 添加
, and four options will pop up:
style: image art style
aspect: output image ratio
negative_prompt: Negative prompt word
controlnet: initial model image
prompt word syntax
There are two main types of grammar: fluent natural language and comma-separated words
This article mainly uses comma-separated words as an example, because this can avoid the unfluent natural language caused by poor English. At the same time, using single words and commas to point out the features you want to see usually has a good effect. We also want to test more parameters, such as adding style, artist style to imitate, physical characteristics, etc., so it is more convenient to adjust with commas.
fine-tuning points
Even if we generate similar or the same prompt words each time, the results of AI drawing may be completely different. If you want to maintain the last result and fine-tune it, you need to keep the seed code of each calculation, which is the number in the figure below sequence
style setting
We said before that you can specify the style in the dialog box, but this can only specify a fixed and relatively broad style, which has relatively large limitations.
For example, we specify an anime style (Anime), but if we want to add a more detailed artistic style, we need to use artwork by
such a keyword, and adding it masterpiece
can keep the work at a high level, which is more effective
# 慕夏的作品,大师级作品,一个女孩,亚洲人,蓝眼睛,沙滩,海洋
/dream prompt:artwork by Alfons Maria Mucha, masterpiece, 1girl, asian, blue eyes, beach, ocean style:Anime
If you deploy the SD Web UI yourself, you can download the model file from the website to change the art style
Recommended website: Civitai | Stable Diffusion models, embeddings, LoRAs and more
prompt word site
SD Models | Live Style: list of artists for SD v1.4 AC / DI / JN / OZ (rentry.org)
Anime Style: Danbooru: Anime Image Board (donmai.us)
Combined prompt word sites:
NovelAI magic guide book (thereisnospon.github.io)
Danbooru Tab Supermarket (novelai.dev)
Encyclopedia of Charms (aitag.top)
Prompt word model example | Download model: [Civitai | Stable Diffusion models, embeddings, LoRAs and more](
reference article
Super detailed! An Introductory Tutorial on Stable Diffusion that Grandma Can Understand - uisdc.com
Stable Diffusion Prompt Teaching | RogicYu's Nest
This article is published by OpenWrite, a multi-post platform for blogging !