Midjourney V4 neural network can generate more specific and realistic images

Midjourney V4 neural network can generate more specific and realistic images

AI image generator Midjourney is available in a new alpha version. The neural network now focuses on more information about creatures or places, recognizes fine details better, and processes more complex clues. V4 can also handle scenes with multiple objects and characters.

David Holtz, CEO of Midjourney, says that V4 has received a completely new codebase and AI architecture. This is the first model trained on the Midjourney AI supercluster, developed over the past nine months.

Holtz describes V4 as the first step towards “something deep and incomprehensible.” He previously predicted a bright future for such neural networks, including AI-assisted video game generation within the next ten years.

Learn important technologies for the tester at hand, and get $1300 already through the robot

REGISTER!

Midjourney V4 can create more realistic images, so the principles of use will become stricter. Anyone who explicitly tries to bypass the list of forbidden words in the tooltip is at risk of being banned.

Overall, V4 will also require a completely new way of wording hints. The remix function, that is, the ability to further edit AI-generated images, is also included in the updated version.

Among the limitations of the alpha version, only a square aspect ratio is currently allowed. In the future, developers will work on improving the resolution and quality of images, adding scale and sharpness settings.

Shortly after the release of V4, Midjourney announced several more updates that focused primarily on the web interface:

  • images can be sorted into collections;
  • collections and highly rated images appear at the top of the profile;
  • profiles can be customized by changing the username and images;
  • widescreen mode, preview and filters added to the interface;
  • added the ability to batch publish / delete;
  • Additionally, you can configure the size of the batch download.

In August 2022, Midjourney had several hundred thousand clients generating millions of images per day across approximately 10,000 servers. Despite the huge size of the project, Midjourney only had about ten employees.

Source: The Decoder

Related Posts

UK to regulate cryptocurrency memes: illegal advertising

Britain’s financial services regulator has issued guidance to financial services companies and social media influencers who create memes about cryptocurrencies and other investments to regulate them amid…

unofficial renders of the Google Pixel 9 and information about the Pixel 9 Pro XL

The whistleblower @OnLeaks and the site 91mobiles presented the renders of the Google Pixel 9 phone. Four images and a 360° video show a black smartphone with…

Embracer to sell Gearbox (Borderlands) to Take-Two (Rockstar and 2K) for $460 million

Embracer continues to sell off assets – the Swedish gaming holding has just confirmed the sale of The Gearbox Entertainment studio to Take-Two Interactive. The sum is…

photo of the new Xbox X console

The eXputer site managed to get a photo of a new modification of the Microsoft Xbox game console. The source reports that it is a white Xbox…

Israel Deploys Massive Facial Recognition Program in Gaza, – The New York Times

The Technology section is powered by Favbet Tech The images are matched against a database of Palestinians with ties to Hamas. According to The New York Times,…

Twitch has banned chest and buttock broadcasts of gameplay

Twitch has updated its community rules and banned the focus of streams on breasts and buttocks. According to the update, starting March 29, “content that focuses on…

Leave a Reply

Your email address will not be published. Required fields are marked *