Users got access to the neural network Midjourney v5 – an updated model that the developers taught to generate more realistic images. The creators explained that the new version uses a different neural architecture, which provides a number of advantages: for example, now the neural network is much less likely to generate “mutations” – a lot of extra fingers or teeth in humans.
As before, Midjourney uses prompts – textual descriptions that allow you to fine-tune your settings. At the moment, Midjourney v5 is at the alpha testing stage, and may receive new features by the release.
Already, the fifth generation has expanded the stylistic range. So it’s much easier to get an image that “imitates” famous artists.
The resolution of the picture was doubled, and the dynamic range was also upgraded. And the image is immediately generated in high quality. But the most interesting thing is that Midjourney now copes with small details, having learned to add five fingers on his hands, and not six or seven, as before.
Earlier it became known about the release of a new language model GPT-4, which received a number of advantages over its predecessor.
