On this website we have often talked about Artificial Intelligence and its practical uses. From AIs that play death metal music , to image filters that make your photos look like drawings straight out of Pixar .
These are just two examples, but they serve perfectly to explain the turn that this type of technology is taking in the artistic field. Some of the tools that currently stand out the most in the field of Artificial Intelligence are those that can create art from written text .
The best Artificial Intelligences to create art from text (or from other images)
AIs that generate images from words require some pretty powerful processing power. For this reason, most applications of this type usually offer limited image sizes in their free version.
In any case, technology has reached a point where we can already obtain truly impressive results, such as images with celebrities, impossible landscapes, dreamlike scenarios and all kinds of abstract art. These are the best AIs to create art from text in 2022.
NightCafe
NightCafe has one of the most powerful AIs to create art and stands out for having more algorithms and configuration options than other similar tools. It’s very easy to use: just write a description of what you want to see on the screen and then choose an art style (oil, synthwave, cyberpunk, 3D game, surrealism, etc.)
After a few seconds, the Artificial Intelligence will display the created image on the screen with a resolution of about 512×512 pixels. In its free version it allows you to make up to 5 images. An interesting detail is that you are the owner of the images created by the AI.
Enter NightCafe
Anonymizer
This tool allows you to create realistic (but artificial) images from other images. The idea is that you can protect your identity online by using a photograph of a person who looks like you, but is not really you.
The operation is very simple: we upload a selfie and in a matter of seconds we will obtain twenty images of people with similar features generated by artificial intelligence.
Enter Anonymizer
DALL-E Mini
One of the most popular art generators of the moment. It is a free and open-source version of DALL-E , an artificial intelligence that learns by reading the descriptions of other images it finds on the internet.
The result may not be as rushed as other machines, but it allows us to make as many attempts as we want. One piece of advice: try to be as specific as possible.
Enter DALL-E Mini
AI Image Generator API (DeepAI)
This tool developed by DeepAI works as a web application, but we can also use it as an API. In general, if we give it a good description, the result will be quite interesting, although do not expect photorealistic images.
Another positive aspect is that the processing time is usually short, so you won’t have to wait long to see the 7 images generated by the algorithm.
Enter the AI Image Generator API
Deep Dream Generator
Another AI capable of generating art is Deep Dream, a tool developed by Google that is turning out to be quite popular. Instead of writing text, what we have to do is upload an image so that the algorithm starts working and generates a new image based on the original one .
The application uses different pictorial styles, many of them hyper-realistic and from different periods and artistic currents. The tool also allows us to choose the category of the image, such as “animals” or “landscapes”, which allows us to obtain better results.
Enter Deep Dream Generator
Starry AI
Unlike other tools, Starry AI is also available in app format for Android and iOS . If you prefer to use its web version, you can also do so, although you must bear in mind that it offers a limited number of credits in its free version.
Starry’s Intelligence is quite versatile, since it has several configuration options, such as the possibility to choose between two algorithms, different image sizes, number of iterations (the more iterations the better results) and image model (realistic or not) .
Enter Starry AI (web version)
BigSleep
Originally developed by @advadnoun , BigSleep is a notebook with which you can create realistic images through a written description or phrase. It is a slightly more complicated tool, since its code is available through Github and we will need a machine where we can run the program (don’t worry, if you enter you will see all the details to use it from the browser).
The app is written in Python and uses a neural network to generate the art. The more iterations we carry out, the better images we will get, being able to achieve resolutions of up to 1024×1024 pixels .
Get into BigSleep (Github project)
Dream by WOMBO
Excellent image generator using AI, available in a web version and also in an app version for Android and iOS . Simply enter a descriptive text for the image and choose a style from over 20 available (realistic, comic, lines, meme, Ghibli, arcane, Dalí, dark fantasy, baroque, etc.)
I personally recommend using the web version for browsers, since in the mobile version they have introduced quite a few payment elements and its use is not as clean or pleasant as it was in the past. In the web version, however, we can generate images and download them without ads or subscriptions.
Enter Dream by WOMBO (web version)
DALL-E 2
This is an advanced version of DALL-E Mini, where its developers, OpenAI, are currently pouring their entire workforce. As they explain, DALL-E 2 is capable of creating illustrations, but it can also design products and even generate new business ideas (if you know how to squeeze its potential correctly, of course).
In this case, Artificial Intelligence can create hyper-realistic images , and it has unique tools such as the brush, with which we can make small strokes to add details such as shadows or enhance certain aspects of the image.
Right now the app is still under development and can only be tested by invitation or by joining the waiting list.
Enter DALL-E 2
VQGAN+CLIP
This other text image generating AI works using the same approach described by @advanoun. VQGAN is an artificial intelligence model that is responsible for compressing and decompressing images based on certain patterns. CLIP is another model whose job is to relate text information with images.
Once both tools are combined, we get an image halfway between the abstract and the hyperrealistic . You can try it by following the indications mentioned in THIS POST .