V3 feedback + idea

[timber flame]

I really like the new improvements from V2 to V3 but the only problem I have is with face reading. and an idea I have is that if the video has someone holding an item you can make the generation detect that object

[misty wing]

@timber flame hi

[terse current]

The official V3 will have fewer facial errors! Plus, we’re working on adding an object detection feature. While it’s not publicly available yet, we’re already using an internal tool to create templates with this capability. It’s set to launch on our mobile app in 1-2 weeks, so stay tuned!

[timber flame]

Oh nice!

[valid salmon]

It would also be cool if it read eyes better. It would make the characters seem so much more human.

[terse current]

I totally agree, the characters could use more emotion! Is there something you do currently to improve that?

[valid salmon]

It helps if the videos you use use intentionally exaggerated facial expressions or if you use well known memes people tend to overlook it anyway.

[terse current]

We will try to improve this in the official V3 version!