Is everyone having a delightful day? Today, after introducing an AI that writes, I’d like to present an AI that sketches. It goes by the name of Craiyon, a blend of the word crayon, typically used for drawing, and AI. It was previously known as DALL-E mini!
1. What is Craiyon?
Craiyon is an artificial intelligence (AI) image generator capable of generating images from text prompts. Craiyon was developed to be a lighter version of OpenAI’s DALL-E and is developed to be as easy to use as the original DALL-E. Formerly known as DALL-E Mini, Craiyon later changed its name to avoid confusion with the DALL-E and DALL-E 2 models from OpenAI, and at the request of OpenAI.
- Direct link to the post about DALL-E >>
The company was founded in 2022 by Boris Dayma and is headquartered in Houston, Texas. The founding came after Boris Dayma took part in the JAX/Flax Community Week in June 2021 and first proposed the concept behind Craiyon on the Hugging Face forums. The initial version was trained for three days on a single TPUv3-8 provided by Google Cloud for the duration of the program. Despite the limitations in time and compute architecture, it was a largely successful concept.
Thumbnails from Craiyon
One of the most astonishing features of Craiyon is, without a doubt, its FREE access!
The Craiyon model underwent an extensive training process with millions of images sourced from the internet along with their corresponding captions. These annotations guide the model in selecting relevant images based on the provided text prompts. It’s designed to merge various concepts, enabling it to craft novel images from a prompt it recognizes.
2. How to use Craiyon?
Most AI-based platforms are developed using their unique technical know-how, but their learning methods through users are generally similar. Let’s dive right in and see how to use Craiyon!
- Direct link : https://www.craiyon.com/
To try out Craiyon AI firsthand, simply head over to the website mentioned above. There’s no need for any separate registration or login process to create art, so it’s easy to get started. Come on and follow along!
Main Page of Craiyon
When you access the Craiyon website, you’ll see an interface like the one shown above, with a familiar text box asking “What do you want to generate?” and a lineup of images under the ‘Inspirations’ section, each showcasing different styles. An interesting feature is that the moment you click on the text box, four options appear at the bottom:
- Art
- Drawing
- Photo
- None
Today, we’ll test and explain each of these four options in detail.
1) Art
In the text box, I tried entering the word ‘Fox’. Following this, I clicked on the ‘Art’ option listed below, and then hit the ‘Draw’ button on the right. The ‘Draw’ button changes its display to ‘Drawing’, and below that, a timer indicates the ongoing learning process.
The waiting time isn’t long, allowing users to quickly receive their generated results. Unlike what was explained in the previous Stable Diffusion article, there’s no setting for batch size, but it does present a total of 9 outputs at once. From all the AI platforms I’ve experimented with so far, it seems Craiyon is the only one to provide this many results in such a short span of time.
I clicked on the eighth image out of the nine results. As expected, there’s an ‘Upscale’ button that allows you to further enhance the selected image, and I immediately gave it a try. For reference, no matter which of the nine results you click on, a window pops up where you can see detailed information about that image (such as image size, the model used “Art”, ratio, Prompt, and Negative Prompt).
Wow! Can you see this stark difference? Do you feel that the face and fur texture of the fox processed initially has become much more delicate and clearer through Upscale? It’s really… incredible. Being able to upscale within the platform without any additional plugins, and accessible for everyone, seems like a really great feature.
2) Drawing
This time, I’ll try out the ‘Drawing’ option, which was the second choice. I’m very curious about the difference between ‘Art’ and ‘Drawing’. To make a precise comparison, I’ll use the same prompt. Please note that I’m not using the ‘Negative Prompt’, so keep that in mind.
You can see that the output is completely different from ‘Art’. It really feels much closer to what you’d expect from ‘Drawing’. Just like before, you can choose an image you like and upscale it, so I recommend giving it a try.
However, please note that perhaps due to the ‘Drawing’ model option, it’s hard to see as stark a difference as you would with the ‘Art’ option.
3) Photo
I will now test the ‘Photo’ model, which seems to be the complete opposite of the ‘Art’ and ‘Drawing’ options. I expect the ‘Photo’ setting to aim for realism. Just like before, I will process it using only the ‘Fox’ as the Positive Prompt
Certainly, compared to the other settings, you can see that the images are depicted based on facts and seem to have a touch of the ‘Fade’ option that’s characteristic of a camera’s output. While there are some unrelated images in the mix, I presume this might be because of the ‘Photo’ setting. If we focus solely on the aspect of naturalism, it feels more real compared to the crispness observed in the ‘Art’ option.
4)None
Now it’s time to try out the last option, the ‘None’ model. Honestly, out of all the options, this one intrigued me the most. Somehow, I felt this option might showcase Craiyon’s true technical prowess… haha. As usual, I used the same prompt.
Oh… this feels fresh. Looking at the overall visuals, it seems to capture realism even better than the ‘Photo’ option. The focus on the subject and the blurred background are distinctly different from what we saw in the ‘Photo’ setting. Of course, perspectives may vary from person to person, but I personally love this option.
3. Summary
So far, I’ve shown you the diverse results from various models just by using the word ‘Fox’. When you all actually use Craiyon, I recommend using more specific sentences or a combination of words, not just a single word. Also, the ‘Negative Prompt’ is crucial for generating the image you desire, so I definitely suggest you try it out.
Advantages :
- Easy interface and usability
- Quick results for the batch size
- Upscaling capability without additional plugins
- Free
Disadvantages :
- Asymmetry in crucial sections
- Issues with detailed proportions
If anyone wants to quickly create their own artwork, I highly recommend Craiyon. However, it seems advisable to go through a lot of trial and error to establish your own workflow process.
I have introduced the TOP 5 AI Art Generators as of 2023. Of course, there are more diverse forms of Generators or solutions than this, and I plan to explain them one by one. Moreover, I will share all the various and trending information to help many people efficiently work using AI services~!