Voice and media
What is AI image generation?
AI image generation creates new images from text prompts using models like diffusion. Learn how it works, how to write good prompts, and its limits.
AI image generation is the use of machine learning models to create new images from a description, usually a short text prompt. You type what you want to see, and the model produces a picture that did not exist before. It's one application of generative AI.
How does AI image generation work?
Most modern image generators are trained on very large collections of images paired with text descriptions. During training, the model learns statistical links between words and visual patterns: what "a red bicycle" tends to look like, how light falls in a "sunset", what a "watercolor" texture looks like.
A common method is the diffusion model. Training teaches the model to remove noise from images that have been progressively corrupted. To generate a new image, it starts from pure random noise and repeatedly removes a little noise, guided by your prompt, until a coherent picture appears. Because the starting noise is random, the same prompt can produce different results each time.
The model doesn't look up or copy a stored picture. It generates new pixels from what it learned, which is also why outputs can contain odd mistakes.
What can you do with it?
- Create illustrations, concept art, avatars and mock-ups from a description.
- Make server banners, event posters or meme templates.
- Edit an existing image: change the style, background or details based on an instruction.
- Explore visual ideas quickly before involving a designer.
How do you write a good image prompt?
Specific beats vague. A useful prompt names the subject, setting, style and mood. Compare:
- Weak: "a cat"
- Better: "a grey tabby cat asleep on a windowsill at golden hour, soft focus, watercolor style"
A few habits help. Describe one clear scene rather than many unrelated ideas. Name a medium or art style if you care about the look. Say what to include, because models respond better to what you ask for than to long lists of exclusions. Then iterate: change one thing at a time and regenerate.
What are the limits of AI image generation?
- Text in images. Lettering is often garbled, though it has improved over time.
- Hands, counts and fine detail. Extra fingers and wrong object counts still appear.
- Consistency. Keeping the same character across several images is hard.
- Accuracy. Generated images aren't evidence. A realistic picture of an event doesn't mean it happened.
- Bias. Models reflect patterns in their training data, which can produce stereotyped results.
Is AI-generated art copyrightable, and who owns it?
This depends on the country and is still being settled by courts and regulators. If you plan to use generated images commercially, read the terms of the tool you used and check the rules where you live rather than assuming.
Common mistakes
- Writing a prompt that is too short and getting generic results.
- Stuffing in contradictory styles, which produces muddy images.
- Trusting a realistic image as a real photograph.
- Posting generated images that depict real people in misleading ways.
Doing this with Based
In servers where the images module is on, /imagine followed by a prompt generates an image in the channel. Based can also create or edit images from requests in chat, and you can attach an image when you mention Based or reply to it to ask about that image. There is a per-member cooldown, with a default of 3600 seconds, which server staff can change. Every slash command also works with the "." prefix, for example .imagine. See the image generation docs and the image generation feature page. For prompt-writing habits that carry over to text too, read how to write good prompts.
Frequently asked questions
How does AI turn text into an image?
Most generators use diffusion: they start from random noise and remove it step by step, guided by your prompt, until an image forms. The model learned the link between words and visuals from many captioned training images.
Why do AI images get hands and text wrong?
The model learns visual patterns statistically rather than understanding anatomy or spelling, so fine structured details are easy to get wrong. Results have improved but still aren't reliable.
Why does the same prompt give different images?
Generation starts from random noise, so each run begins differently. Small changes to the prompt also shift the result.
How do I generate an image with Based?
Use /imagine with your prompt, or ask in chat by mentioning Based. The images module must be enabled for the server, and a per-member cooldown applies.
Keep learning
- AI basics
What is generative AI?
Generative AI creates new content such as text, images, audio and code from a prompt. Learn how it works, what it can make and where it falls short.
- AI basics
How to write good AI prompts
A good AI prompt states the task, context, format and constraints clearly. Learn practical steps, examples and common mistakes to get better answers.
- AI basics
What is machine learning?
Machine learning is a way to build software that learns patterns from data instead of following fixed rules. Learn how training works and where it is used.
- AI basics
What is an AI chatbot?
An AI chatbot is software you talk to in everyday language, powered by a language model. Learn how it works, how it differs from older bots, and its limits.
Try it in your server
Based is an AI Discord bot in 300+ servers. Every slash command also works with the "." prefix.