0800-31-0700 for new subscribers
0800-31-0800 technical support

Testing popular neural networks for image generation

Home /

Blog

/

Testing popular neural networks for image generation

Testing popular neural networks for image generation

22.12.2023

Internet

11180

Neural networks can be used to generate not only texts but also images based on the entered text. The results are available almost instantly and depend on the capabilities of the software. We decided to test five popular free AI generators to make our own impression of them and share the results with you.

Bing Image Creator

This service for creating drawings appeared in free access in autumn 2022. Today, users generate more than 2 million images every day with its help. So far, Bing Image Creator supports only English, but others are already in development.

Bing Image Creator is powered by artificial intelligence powered by the DALL-E neural network and produces results just as quickly. It is accessible via Bing Chat and browser. You'll need a Microsoft account to log in. Conveniently, Bing Image Creator can be accessed in the same place as the Bing chatbot. Once you have access to Bing Chat, all you need to do is put it in creative mode and give it a text prompt. After that you get the image you want. In the future, it is planned that the service will also be available in clear and balanced modes.

We share with you the result we got from using Bing Image Creator.


Deep Dream Generator

The online service allows you to easily turn photos or text into real masterpieces in different styles. Basic functions of the programme are available in the free version, and advanced ones – in paid packages. In the free version, the user is provided with 100 units of energy that are spent on queries, which is quite enough to understand the generator's capabilities.

Deep Dream Creator offers three main modes for working:

  • Text2Dream – image creation by text description;
  • Deep Dream – deep processing of uploaded images with customisation of parameters;
  • Deep Style – image stylisation for certain samples.

We chose Text2Dream mode for the experiment and also asked to draw a computer of the future. We got this picture.


Stable Diffusion

This image generator from Stability AI became available to users in autumn 2022. The completely free neural network in 2-3 minutes generates four variants of 512x512 pixel images in a chosen style based on any text description. On the "Advanced options" you can reduce the number of images and also adjust some other options.

Interestingly, from the very beginning the programme was capable of creating pictures "in nude style" and with personalised content. After the algorithm update, this capability was limited.

Here is one of the AI-drawings obtained by the request "create an image of the computer of the future" in cinematic-style with Stable Diffusion.


Dream by WOMBO

Neural network allows generating pictures on Android and iOS devices. To do this, you only need to come up with a quality prompt – a command in the form of text up to 200 characters long with a description of the desired picture. After that, choose a visualisation from the offered options. As a result, you will get an image with a resolution of 960x1568 pixels, which you can download or share on social networks.

Also Dream by WOMBO service is equipped with an editing function, which allows you to refine the details of the picture with the help of text hints. In the free version, a picture can only be edited twice, and then you have to upload a new request.

The free package gives you access to the basic features of the programme. If you need advanced features (priority in the generation queue, high resolution, ability to make more edits), you can subscribe to the programme for a fee.

Here is the result of using the free version of Dream by WOMBO.


Leonardo AI

Although this service improves image detail worse than the paid Midjourney neural network, you can use it for free. Leonardo AI provides four variants of pictures at your request.

The neural network has a built-in prompt generator, a gallery of photos, and also gives you the opportunity to generate a picture according to the provided photo. But it doesn't yet know how to create variations of images according to the already generated one (for that you need to enter a new command). The programme is available in a full web version and a mobile app.

Below is the computer of the future "through the eyes" of Leonardo AI.


Conclusions

When choosing between different options of image generators, consider such criteria as usability, quality of results, speed and affordability. Then you will have a really useful tool in your hands that will help speed up and simplify your work.

Comments

1

DM

20.03.2024

Image Creator від Майкрософт сподобався, непогані роботи видає + безкоштовних запитів достатньо

Макснет

20.03.2024

Дякуємо, DM, що поділились своїм досвідом.

Еще комментарии