
Had a chance to play around with the mid journey video feature and it is quite amazing.
Prompt Image:
Animation generation didn’t really take any effort. Just type in some text and let it generate.
Another example:
Prompt Image
It also can gerenate near flawless 3D camera movements.
Promot Image:
https://www.youtube.com/watch?v=3klVeNEb6vY
Prompt Image:
Alternate version with pan.
Notes:
- 3D understanding on these scenes are phenomenal. We can probably extract out 3D map, then export the semantics out from it.
- We can probably ask Midjourney to just segment out some of the sections. → So far couldn’t find a way.
- I feel like without even working on those semantic engines and what not, we can make a full world using generative AI.
- The semantics will come eventually but for now manual markings may be the way to go if we want speed.
Comments
Join the discussionNo comments yet. Be the first to comment.
Machi
Follow the development of Machi, a side-scrolling simulation world built to test AI agents, tile-based emergence, and the future of embodied intelligence. Coming Soon: https://machi.sprited.app



