How the neural network sees me: 6 services that remake photos

1. Stable Diffusion XL

The popular neural network Stable Diffusion XL is able to complete finished pictures, expand the boundaries with new details, and also convert one image to another according to the described request. The last function can be used to process portraits.

You can test Stable Diffusion XL online when working with photos through the Clipdrop service. It has a section for converting Reimagine images. The neural network will analyze the uploaded portrait and offer three variations with other angles and clothing. However, the tool does not allow detailed customization of the result. But all generated versions usually retain the general color scheme and shapes of objects.

Try Stable Diffusion Reimagine →

2. Kandinsky

The Sber neural network is capable of generating variations of finished images and combining styles from different frames. Kandinsky supports several operating modes. In standard, the service creates a picture only based on the entered text message.

To transform portraits, the mode of combining two frames into one new one is useful. You can also upload a photo and enter a query with additional characteristics. In this case, Kandinsky will generate a new frame, taking into account the proposed file and the prompt at the same time. This function currently only works through the official Kandinsky bot on Telegram. Sometimes the results are strange or funny - it depends on the original image and the specified parameters.

Try Kandinsky 2.2 →

3. Prisma

A popular neural network-based application for smartphones and browsers. It allows you to process any images, including portraits, in different styles. For example, you can choose specific visual references, such as famous artists or illustration methods.

Using the mobile version of Prisma, you can improve photos - remove objects, change contrast, background and lighting. At the same time, the system offers suitable functions for each photo after scanning. For example, for a selfie, the service will advise you to treat your eyes, eyebrows or lips.

To use all the features, you will have to subscribe - a month costs $5.

Try Prisma Web →

4. Runway ML

Runway ML is a large set of neural network-based tools that can be trained and used for various tasks. Among the functions there is also image conversion. The service can process portraits and generate new versions of pictures in a whole series of different styles. The system produces generally good results, although much depends on the contents of the frame.

Finished files can be exported and saved to a computer. In free mode, only 768 × 768 pixel resolution is available. High quality files are unlocked with a paid subscription starting at $15 per month.

Try Runway →

5. Fotor GoArt

Fotor - popular photo editor, which has neural network-based tools for working with images. First you need to upload a picture, and then process it in one of the selected styles and add the desired effects.

Due to its ease of use, the service is well suited for novice users. If you don’t like the final result, you can repeat the generation several times. In the free mode, you can take several images, but for full access and saving pictures in high quality, you will have to sign up for a subscription costing from 690 rubles per month.

Try GoArt →

6. Deep Dream Generator

Deep Dream Generator has been around for several years now. This is a popular picture generator with additional features for converting ready-made images. It uses Google's neural network, trained on millions of examples.

The service is easy to use; all you have to do is upload a portrait and select a processing option. The system will automatically generate a new image based on the original one.

There are three operating modes available in the Generate section: Text 2 Dream, Deep Style and Deep Dream. The first one is designed to generate a picture by text request, in the second you can specify a specific artistic style, and the third is needed to create experimental psychedelic images produced by the Deep Dream neural network itself based on downloaded source.

Try Deep Dream Generator →