How to use Craiyon, Artificial Intelligence that creates any image for you

Every day is a new opportunity for you to find something new — and maybe fun — on the internet. Well, today you will discover a site where you can spend some time having fun: the crayon. When accessing it you are faced with a space to type something (literally anything) and the artificial intelligence of the system offers you 9 different images according to the words offered. Learn how to use it and check out some tips about the feature!

Craiyon, the designer website

One of the easiest ways to use the crayon it's entering what you want to see drawn by artificial intelligence along with some existing art style. Junctions like "Power Rangers in the style of Andy Warhol","Elden Ring by Katsushika Hokusai" or "cave painting of a giant squid” might be some valid suggestions — the translations might be something like “Andy Warhol-style Power Rangers","Elden Ring by Katsushika Hokusai"and "giant squid painted in a cave", Respectively.

Despite having been searched in English, the platform can also recognize some requests in Local Guide. But since it has been trained in the English language, artificial intelligence will be a little more successful when researched in this way. When isolated or simpler words are inserted, the results are usually more accurate too — but the cool thing is that you can explore the system's capabilities to recognize exactly what you want to see.

As it is a search that takes at least three minutes according to the website itself – but in practice it is much less than that – you may not get exactly the result you have in mind, but the surrealist pattern presented may impress you. .

A curiosity is about its origin. Craiyon was based on the DALL-E Mini, an aspect of GIVE HER, AI that makes exactly the proposal of the site, to present images according to the searches of the users. The name is a portmanteau of Wall-E, from Disney, with Salvador Dali, something metaphorical to associate technology with art. Currently, to access DALL-E it is necessary to enter a queue, so to avoid it, let's take advantage of another tool that works in a similar way. See how Craiyon works.

How to use Crayon

1. The way to use the platform is very simple and intuitive. First, go to Craiyon website;

2. After I searched for what you want the platform to design for you. researches in English may be a little more assertive than surveys in Portuguese, but even in Local Guide the system delivers some interesting—or even expected—results;

Wait a moment and that's it! Just below the result you can click on “Screenshot” to have a screenshot downloaded to your device of the result or click on “Forum“, to open the Hugging Face page and be able to share your result on this specific social network.

Tips for using Craiyon

Because it is a platform that has been making a lot of buzz on the internet, many people want to make the most of the feature. Here are some suggestions on how you can carry out searches that have results closer to what you want — or at least that you know how the platform works to have more creative images.

Things in specific places or describing actions

You can make a wider request by asking Craiyon to place objects in specific places. Try something like “the sun in a jar” or “the Demogorgon on the Titanic”, and you should get at least one image that resembles what you want. Other actions like “Mickey holding a candle" or "Darth Vader with an umbrella” are very specific searches that can work.

messing with food

Craiyon can usually understand when we refer to food, with a few odd exceptions. "Meatloaf” worked fine, as did “chocolate sushi”. But artificial intelligence doesn't seem to understand what ramen is, or sometimes just noodles. Both the searchhumpback whale jumping from ramen" as for "humpback whale jumping from noodles” we only had image results of a whale coming out of the ocean normally. No noodles came with the animal, just salt water.

Be specific, but not too much

Sometimes you can get lucky with some well-crafted requests, but it seems like the AI ​​might run out of space to describe all things. We get very accurate image results, such as “a bear pretending to be a high school student, wearing a red backpack”, but once we said that the teacher was “and writing on a blackboard while a teacher is around”, the result failed to show the bear doing his schoolwork.

Details can help

Including some finishing or quality characteristics like “detailed”, “4K”, “8K” or “photorealistic” may not provide even more detailed image results. We tried several times but didn't notice much difference. "Kanye West, marble sculpture"and "Kanye West, marble sculpture, detailed” and they both looked pretty much the same. What worked a little better was adding “anatomically correct” to searches. (https://theshabazzcenter.org/)

What doesn't work – for now

The site is still learning how to use it, but for now it doesn't seem to understand the denials. For example, when asking to spawn a specific character without a head, all heads will remain. This is because the program's developers consider it inappropriate to use its AI to create images that people would find disturbing, distressing or offensive. To try "Statue of Liberty without a torch” you can't get the monument result without the object she holds, but the result delivers something similar to cutting the torch from the images.

And you, did you get cool combinations? Tell us in the comments!

See also other features

Learn About Spotify Pie, which shows your most listened music genres and artists.

Pig iron: PopScience.