Profile Picture

Proposal of AI functions that are popular in Japan

Posted By yoshihi 2 Months Ago
You don't have permission to rate!
Author
Message
yoshihi
Idea.png
yoshihi
Posted 2 Months Ago
View Quick Profile
New Member

New Member (55 reputation)New Member (55 reputation)New Member (55 reputation)New Member (55 reputation)New Member (55 reputation)New Member (55 reputation)New Member (55 reputation)New Member (55 reputation)New Member (55 reputation)

Group: Forum Members
Last Active: 2 Months Ago
Posts: 2, Visits: 27
In Japan, creating video content using AI is very popular, especially among junior high and high school students. If iclone had a casual function that allows you to create simple videos using its AI function, I think even more Japanese people would use it.
Here are some suggestions for the new AI functions in iclone that Japanese people think about.

A function that automatically generates movements and actions that match your avatar's personality by inputting it in text.
For example, if you enter "drunken man", an avatar of a drunk man will be generated.

Crowd control has a function that automatically generates a crowd when you enter the situation in text.
In war scenes, automatically generates soldiers' behavior such as firing guns and falling down when shot.

A function that automatically constructs a 3D map of the cityscape by entering text into the map editor
For example, if you enter "New York-like cityscape", a 3D set with a New York atmosphere will be automatically generated.

A function that automatically performs camera tracking and camera effects by inputting text.

A function that automatically generates an animation in which the avatar speaks the same words as the creator speaks into the microphone.
AI converts voice quality, gender, age, etc.


I think it would be good to have an AI subscription option that allows you to use those AI functions.
Sophus
Sophus
Posted 2 Months Ago
View Quick Profile
Veteran Member

Veteran Member (746 reputation)Veteran Member (746 reputation)Veteran Member (746 reputation)Veteran Member (746 reputation)Veteran Member (746 reputation)Veteran Member (746 reputation)Veteran Member (746 reputation)Veteran Member (746 reputation)Veteran Member (746 reputation)

Group: Forum Members
Last Active: 1 hour ago
Posts: 202, Visits: 2.3K
I don't think that will be very easy to do. It's highly resource and cost intensive to train AI models.

I believe OpenAI pays about 7 BILLION(!) Dollars for training of AI models alone (wages for employees aren't included in that number) and are nearly bankrupt by now. And it's a much smaller company than other AI ventures which are financed by Google or Microsoft or Apple or Meta.

This can only be done with the financial help of big cooperations or investors.

There are certain AI models out there which can convert images to primitive 3D models, for example TripoSR or Dust3r. But at the current state it's more of a proof-of-concept-phase. Will probably take some more years of research and training.



Reading This Topic