Midjourney v5 is the latest language model of the popular text-to-image generator known for its realistic creations.
The update rolled out to Midjourney’s paid customer base on Wednesday and many users, including Graphic designer Julie Wieland, have been sharing their new AI-generated artwork. AI details that the v5 language model brings with it include improved “efficiency, coherency, and quality,” Midjourney said on its website.
Many users have especially noted the upgrades in the human hand details, with subjects more likely to have five fingers per hand. In previous models, human hands were often generated with an incorrect number of digits, anywhere from four to between seven and ten. Those who have been able to test Midjourney v5 have been impressed with the visual upgrades.
“The latest Midjourney v5 model is both extremely overwhelming/scary and beyond fascinating,” Wieland told PetaPixel.
“Its ability to recreate intricate details and textures, such as realistic skin texture/facial features and lighting, is unparalleled,” she added.
While Midjourney is a somewhat unique language model, it works in a similar fashion to other text-to-image generators such as Stable Diffusion and DALL-E, using worded prompt descriptors to create images, as per the model, which is trained using human-made art as a reference, according to ArsTechnica.
There has been an ongoing debate about whether AI-generated art is legal or ethical; however, many Midjourney v5 users consider with continued upgrades it might be indistinguishable from real art and photography. Wieland considers that AI image generators, such as Midjourney are likely to be companions for photographers rather than replacements.
However, Midjourney v5 isn’t without its flubs. AI image creator Nick St. Pierre shared many of his works, for which he used the same prompts in the previous Midjourney v4 and Midjourney v5. One image was a fascinating shot of young women in 1960s street style that resembles an outdoor fashion shoot. Upon closer inspection, you can spot the floating hands not attached to bodies and misshapen faces in the background.
With its latest v5 model, Midjourney founder David Holz recommended on the brand’s Discord that users shy away from short prompts in favor of “longer, more explicit text” to get better results.
- AI-generated videos have arrived, and they’re evolving fast
- Here’s how ChatGPT could solve its major plagiarism problem
- The best AI image generators to create art from text
- AI image generators appear to propagate gender and race stereotypes
- Meta made DALL-E for video, and it’s both creepy and amazing