Midjourney 5.1 has arrived. It’s more “opinionated” in its image generation but also comes with a “raw” mode that removes that trait and is sharper and more accurate, with fewer unwanted borders or text artefacts in images. During testing, Forbes’ Barry Collins used the same prompts in version 5 and version 5.1, left all other settings the same and everything set to base quality and medium style.
In the image above, Midjourney 5.1 overcomes a persistent problem with the previous version – over-aggressive, artificial-looking lens flare – to produce images that are more natural, with warmer, more interesting skies.
Again, the version 5.1 images much stronger than the version 5 equivalents; more colourful, with more dynamic compositions.
Here, version 5.1’s “opinionated” trait comes through. While the version 5 images on the left look like snapshots, the 5.1 images on the right look like they might have been taken by a professional photographer.
While both versions misunderstood the prompt, delivering what looks like a salesman with a newspaper rather than a newspaper salesman, with text in the background, rather than the newspaper itself, there’s little difference between the two versions’ output; he text in both sets of images looks OK from a distance, but when you zoom in close it’s largely non-sensical. The version 5.1 images have stronger compositions, with the subject filling more of the frame.
Midjourney 5.1 is expected to be made generally available as the default system as soon as the end of this week.
If you are looking for advice, guidance and support on safely using generative AI tools and other forms of machine learning for content creation, including to optimise and improve processes and workflows, get in touch via the Mondatum Contact Us page.
Source: Forbes