Machine learning and neural networks can be cool tools for photographers and artists, but integrating them into your workflow is a challenge for those who can’t code. We have put together a selection of applications based on neural networks that are easy to understand and start working in just a couple of clicks.
A free online service in which a neural network improves the quality of a photo. It removes noise and increases the resolution of images by the desired number of times, and also sharpens. Extremely easy to use interface.
Let’s Enhance is a commercial analogue of Waifu2x. The purpose of the service is the same — it is a neural network to improve photos online. You can increase the resolution of the picture by 16 times, as well as adjust its color and tone and remove noise. You can upload several images to the site at once and start processing them simultaneously.
The free version allows you to increase the resolution by 4 times for only five photos. The subscription allows you to increase the resolution by 16 times for 100 images and costs $9 per month. For processing 300 photos, you will have to pay $24, and for 500 — $34.
An image generation service with the following functions:
1. “General” (General) — allows you to cross any image with each other. Nothing prevents you from crossing a pizza and a parrot, but the result will be abstract and far from photorealism.
2. “Portraits” (Portraits) — allows you to collect the portrait you need from several images of faces. Faces have customizable options. You can set the physique, age, gender, mood, realistic/drawn facial outlines, and so on.
3. “Album Covers” (Albums) is the generation of abstract images in the manner of music album covers.
There are also functions for generating landscapes, paintings, buildings and animated characters.
In order to use Artbreeder, you will have to stand in line or pay a tariff, the cheapest of which costs about 700 rubles. The paid plan will allow you to save the generation results to your computer at a resolution of 2048x2048 (in the free plan, the resolution is 4 times lower, and you can only save the first 4 generated images), as well as generate images based on those that you upload to the platform.
Helps you retouch your photo with the smart brush. With the help of Nvidia InPainting, you can automate the design routine of removing unwanted objects from a photo. So, if visual garbage got into the frame or you need to remove an object sticking out of your pocket or correct skin defects, the application will cope with everything. The algorithm analyzes the pixels located around the selected object and tries to replace them in the most appropriate way.
If in the previous service we removed objects from a photo, now we remove the background around a certain object. There is a similar function in Adobe Photoshop, the Adobe Sensei neural network is responsible for it — we wrote more about this in a separate article.
There are versions for Windows / MacOS / Linux, plugins for Adobe Photoshop, Figma and other programs. The free version of the application greatly limits the resolution of the downloaded result, making it unsuitable for professional use (unless, of course, you are going to increase the resolution of one of the free neural networks mentioned above in the next step). The cost of the cheapest subscription is 579 rubles. For this price, the service allows you to edit and download high-resolution results for 40 photos.
An algorithm that details blurry faces in a photo. Relevant for cases where the shooting took place in poor lighting or “on the go”. The algorithm recovers well even very compressed images. Although it happens that it draws unnecessary details or makes the image too cartoony. You can get acquainted with the work of the algorithm by the link.
Once you have followed the link, place the image you want into the GFP-GAN workspace. Then click Submit and wait about 5 seconds for the algorithm to complete its work. The result and the original will appear in one window.
Application for Android and iOS (there is also a web version) that allows you to transfer the style of famous artists to photos. And this does not happen by applying filters. The image is analyzed by a neural network, after which it is completely redrawn.
The service colors black and white photos in realistic colors.
This is a shareware service. Each illustration will have the inscription “Created by colorize.cc”, and you can upload up to 50 pictures in total. To remove the inscription and increase the download limit to 10,000, you need to pay $10.
An application that combines many of the neural networks that we have already talked about and adds others. Here you can generate images, remove objects from photos and videos, create text captions for objects in pictures, create a 3D model from photos, and much more. All calculations take place in the cloud. The application has a web version, as well as versions for Windows and Mac.
Runway ML is free to use, but there will be restrictions on uploading your work in high resolution and using some tools. The cheapest plan without these restrictions costs $35.
Time to learn tools that speed up processing and offer new creative techniques. The main thing to remember about the features of AI-based algorithms is that the result is not always completely under the control of the user. Therefore, classic graphic editors do not become a thing of the past, but work in synergy with new tools.