[10:20 Tue,2.April 2024 by Rudi Schmidts] |
Midjourney plans to launch the seventh version of its popular AI image generator within the next three months, which will feature a new function: Before the summer, Midjourney v7 will then offer "personalized AI models" by default.
In model personalization, the AI uses the user&s individual preferences instead. These user preferences are in turn captured through image ratings. Personalization thus overwrites the standard behaviour of the model. Explicit instructions in the prompt are also prioritized so that the personalization only affects the elements not specified in the prompt. For the current version 6, Midjourney will probably deliver a further update to improve bodies and hands before then, and a possible speed update is also still under discussion. Holz also announced that Midjourney plans to release a video model before the end of the year. The previously announced 3D model is also likely to be inserted before the "video update". ![]() It is an open secret that Midjourney is also working on a video model. But the fact that the date "before the end of the year" has now also been nailed down raises a few questions. Will the model be so good that the field can be left to the www.slashcam.de/news/single/Revolutionaere-Video-KI-Sora-soll-noch-dieses-Jahr--18461.html (current SOTA model (Sora from Open AI)) for the next 8 months without a fight? Or does Midjourney simply not have enough computing capacity for training to realistically be able to finish a first version before then? It is also a foregone conclusion that other major players such as Alphabet/Google want to present new video models this year that will compete with Sora. Quite a few observers assume that 2024 will be the year for AI video. ![]() deutsche Version dieser Seite: KI-Generator Midjourney v7: personalisierte KI-Modelle, anschließend 3D, dann Video |
![]() |