In this article, we are going to show you a fantastic program called Craiyon. It is an AI system that can generate any image from a text prompt. The images you can create are literally only limited by your imagination. It’s a great way to learn about some of the possibilities of AI while having a little fun at the same time.
In this article, I’ll explain the basics of Craiyon, how to access and use it, and some creative prompts to get you started creating your own fantastical images.
What is Craiyon?
Formerly named Dall-E Mini, Craiyon is a user-friendly offshoot of OpenAI’s much more advanced DALL-E image system. As of writing, DALL-E is not readily available to the public. But Craiyon was developed as a free program open to everyone, based on the same idea.
Craiyon is an AI model, or neural network, that converts plain-language text to images. It is not the most powerful image AI on the market at the moment, but it’s probably the most user friendly.
What can Craiyon do?
Ever wondered what a dog with a giraffe neck would look like? Wonder no more…
Craiyon has become the world’s favorite meme image generator, but it’s much more than that.
You can create any image and dictate in what context, from what angle, and in what style of art the image is generated. From oil painting, pencil drawing, and line art, to impressionism and spray-painted graffiti, anything is possible. You can depict your subject as a futuristic cybernetic robot or as a stained glass window.
If you have a favorite artist, you can create an image in that artist’s style.
How To Use Craiyon
The beauty of Craiyon is its ease of use. Just go to www.craiyon.com and click inside the text box at the top and describe the image you would like to see.
Craiyon will display nine images and it allows you to take a screenshot of all nine. You can also click on any one of the nine images to zoom in and either screenshot or download the created image.
Play around with it a bit, you’ll likely be able to generate some funny and creative images almost immediately. But to really make impressive work, you need to know how to describe the images you want Craiyon to create.
The key to getting the best images from Craiyon is to be specific with your prompts. As fun to play around with and as amusing as the images it can render, Craiyon actually serves as a great tool to fine-tune your searching lingo. After all, “prompt engineering” is likely to become a future job qualification (it may already be).
The key to getting the best images from Craiyon is to be specific with your prompts.
These tools only really work if you know how to prompt them correctly. So keep reading and we are going to give you 10 great tips to help you refine your search prompts.
10 Tips for Fine-Tuning your Image Search Prompts
Craiyon will only be as good as your prompts. To get a truly amazing images, you’ll need to be as specific as you can about telling it what you are looking for. We are going to go over 10 tips to fine tune your keyword prompts and show you how you can get the best results.
1. Try It Again
Craiyon is free and there are no limits, so have at it. If you don’t like the first nine results it gives you, hit the “Draw” button again until you get an image you like.
2. Try Huggingface
Go to https://huggingface.co/spaces/dalle-mini/dalle-mini and use the same program which will garner different results. Craiyon (formerly Dall-E Mini) created by Boris Dayma, a Texas-based computer engineer, was a result of a program developer competition hosted by Hugging Face in conjunction with Google in which Dall-E Mini was introduced.
The name has since been changed to Craiyon and is hosted on its own site. It is constantly being tweaked and improved on the new Craiyon site, so the same program on these different sites can give you different results. So if you don’t like what you’re getting on craiyon.com, try the same keywords on huggingface.co.
Below are two images using the prompt “steampunk dog art” – the first on www.craiyon.com and the second at https://huggingface.co/spaces/dalle-mini/dalle-mini. You can see how the image results are slightly different.
3. Add An Action or Location
The prompt doesn’t have to be just one word. If you’re looking for an image of “X”, try things like X jumping over the moon, X on a swing, X in jar, X caught on night vision trail cam, X surfing on a lava flow. You can also try things like on a rooftop, in the desert, peeking out of a burrito/pile of laundry/from behind a curtain. If you don’t give the image subject a location reference, you’ll likely end up with something on a white background.
4. Add An Art Style
You can add “in the style of”, or “by” and add an artist’s name, like Pablo Picasso, Rembrandt, Claude Monet, Salvador Dali and it will create an image in the painting style of that artist. You can use any famous artist’s name. As long as it’s famous enough, you should get some interesting results.
You can also try a style of art like Japanese woodblock, impressionist, or cubist. Or try an artistic medium like watercolor, line drawing, pencil sketch, steampunk art, cybernetic, concept art, or spray-painted graffiti. Try using an art style specific to a certain culture, like Australian Aborigine art or Indian art.
5. Add texture
Made of…. satin, glass, jelly, plastic. Ever wonder what a chair made of porcupine quills or dog face made of cheese would look like?
6. Type of Image/Photography
Try adding terms like scientific illustration, movie poster, or book/album cover to your prompt. If you are looking for true-to-life photographic images, add modifiers like editorial, magazine photo shoot, studio photography, trending on ArtStation, photorealistic, detailed, nature photography, or National Geographic to improve the quality.
7. Specify the View
You can specify the angle at which you want to view the photo. Type in aerial view, long shot, portrait, wide-angle, fish-eye lens, extreme close-up, macro, from behind or worm’s eye view.
8. Bokeh
This is a feature that earns itself a separate heading. You know all of those cool background images that are just a little blurred out?
That’s bokeh.
So whether you are looking for an image that is an urban cityscape at night, or a rural field of random blooming flowers to use as a softly blurred background, bokeh is your magic keyword.
9. Set The Mood
Try using adjectives like dark, colorful, futuristic, vintage, abstract, soothing, retro, foggy, early morning, funny, happy, depressing… whatever mood or feeling you want your image to convey.
10. Use Lots of Prompts
And lastly, use lots of prompts. Separate keywords and phrases with a comma. The more specific you are, the better the result.
What Craiyon Has a Hard Time With
As amazing of a tool as an AI image rendering program is, you needn’t worry that Craiyon (or any of the other AI image programs) is going to put designers or stock photo sites out of business. Craiyon is not without its limitations.
1. Written Text
Craiyon cannot recreate written text. When prompted with ‘A blue ceramic coffee mug with the words “World’s Greatest Dad” written on it’, this was the output:
2. Faces and Hands
The program has a hard time with faces. They are often distorted and sometimes unrecognizable. It can also come out with images of animals or people with extra or missing limbs or appendages. It really has trouble with hands, in particular.
Looking for “woman holding her hand up in front of her face to show off her new diamond engagement ring” might get you images like this, with too many (or not enough) fingers and distorted features.
There are online threads and memes solely dedicated to the amusingly inaccurate images that come out of AI image generators.
It should be noted that it’s not only Craiyon that has a hard time with hands. All of this generation of AI image-generating software seems to have issues with them.
Take Away
By now you will have realized that the AI is only as smart as you are. It can’t tell what you’re thinking, and can only give it what you ask for as best as it can.
Craiyon is by far the most easily-accessible and easiest-to-use program of its kind out there at the moment and a great demonstration of how to use AI to generate images. It can be a good option to train your input design (prompt engineering) skills because you can clearly see how radically the quality changes with better, more specific prompts. Better input – better output.
Will this replace artists and graphic designers? No. It takes a lot of time to get images you like and the image sizes are very small with a low resolution. It’s great for making memes, or using it as a mock-up for ideas to present to a client, but the image quality is not good enough for professional work.
In its current state, celebrities and politicians also need not worry there will be deepfake images made of them in compromising situations as it’s simply not advanced enough to recreate faces well, especially those of specific people, to be any sort of a threat… for now.
And although it may take a while to find the exact image that you’re looking for as you refine your prompting skills, it’s a lot of fun to play around with and a great way to ease yourself into this new AI era we now find ourselves in.
