Midjourney is one of the most advanced and photorealistic artificial intelligence image generators available today and will be even better when version 6.5 is released. This new update improves realism and enhances skin textures.

According to a post on X by AI art expert Nick St. Pierre, Midjourney CEO David Holtz said in his most recent "office hours" that work on the new update is in progress.

Holtz said on the call that v6.5 will improve image clarity, as well as certain areas such as hand quality and skin texture. In addition to model upgrades, web apps and personalization will also be improved.

AI artist Alie Jules also attended X and reported that the first release will not be widely available, but will be available as an alpha version for some users and will be stress tested. This is the same as past releases and is a common approach for AI companies.

Each new update to Midjourney brings changes to specific features and models that were not available in previous versions. Version 6, for example, improves text rendering and significantly improves photorealism.

In 6.5, we will see improvements in the quality and density of certain pixels and updates to the way skin and facial tones are rendered. This will allow us to create even more realistic humans; Midjourney is already better than any model I have tried.

In previous office hours, Holtz suggested the possibility of splitting the next iteration of updates into two, the first focusing on pixel quality and the second on hand and body cohesion, but it now appears they are working on one for the 6.5 release.

Other updates likely to take place in 6.5 are improvements to text rendering. While this was improved in v6, it is not as accurate as other models such as Ideogram, Leonardo's Phoenix, or DALL-E 3.

As well as the 6.5 progress update, Holz also confirmed that they are still working on version 7.

Personalization was introduced earlier this year to allow users to customize how images are generated based on their preferences.

Holtz also confirmed that they have not started final training for v7 and that testing and other work needs to be done before release.

Other updates include the fact that they are behind on building a style exploration tool, but may be working on a simpler solution to give people more style ideas more quickly available within Midjourney.

The update also confirmed that 3D and video models are still under development, but will take longer, especially for video, as Midjourney is taking a more detailed approach that allows for better control and realism than the current trans-diffusion model may take longer.