Midjourney starts training video models in January, v6 updates coming soon

Midjourney will start training video models in January. This was announced by Midjourney CEO David Holz in an "Office Hour" Discord session. Holz already mentioned in the past that Midjourney is also working on 3D and video generation. It will be interesting to see if Midjourney can gain a qualitative edge in video generation because it builds on a mature image model. Meta has achieved something similar with Emu Video, which is based on the Emu image model and performs better in user preference tests than other current video models. Further updates to Midjourney v6 with improved text rendering and prompt following will be released starting next week. Features from v5, such as inpainting, will also be gradually added to v6.