Midjourney is one of the most popular AI art generators, with the ability to create a wide range of artwork and images. The service is now testing a V5 update with many improvements.
Midjourney announced an alpha release of the V5 algorithm, which is available for testing by appending "--v 5" after an image prompt, or selecting V5 from the settings. According to the announcement, it supports a "Much wider stylistic range and more responsive to prompting," more detailed images, and twice the image resolution of the previous version.
The post explains, "Today's test is basically a 'pro' mode of the model. It's MUCH more 'unopinionated' than v3 and v4, and is tuned to provide a wide diversity of outputs and to be very responsive to your inputs. The tradeoff here is that it may be harder to use. Short prompts may not work as well. You should try to write longer, more explicit text about what you want."
There are also a few new experimental options in V5. You can use the "--tile" argument for seamless tiling, which should work like Tileable Stable Diffusion and other similar tools. The "--ar" argument can be used to create images with aspect ratios greater than 2:1, and "--iw" can weigh image prompts versus text prompts.
Nick St. Pierre shared some demonstration images in a Twitter thread, and the results look impressive. Generated artwork of people has less of the expected Midjourney look (faded backgrounds, too sharp contrast, etc.) and more like real-world photos. The update also seems to be better at creating hands with the correct amount of fingers -- something that previous Midjourney versions struggled to create.
The announcement also mentions that the new model will be "significantly modified" as development continues, so your results with a certain prompt might be different in a few weeks or months than what it now creates. Even after V5 is complete, the older versions should remain available, so you'll still be able to access the current stable model if you prefer that.
Source: Reddit, Twitter (@nickfloats)