Each clip in the video is about a second long. All current video generation models start showing their weirdnesses when there is too much movement. None of these can keep that impression of consistency for longer times. I can spot problems in almost all of the clips in this teaser (robot leg is the most obvious one).
Is there a concise industry-term for "amazing demo, terrible irl"?
Not saying this is one (the demo looks great). But the last 12 months of any AI/ML tools has been 1) demand powered by an amazing, over-hyped demo followed by 2) incredibly lackluster usage as people realize it was overpromised.
You can argue any product as a gap b/w expectation and reality, but the last year has taken it to an absurdity.
The demo video looks amazing. Of course, every AI demo looks like it's the most amazing thing that's ever happened. I eagerly await getting off the waitlist so that I can try it out myself.
What's the underlying model? Do they have their own? Or are they using stable diffusion/video diffusion?
Wow are the sounds also generated with the video? Did not expect something like this so soon. If I were in the film industry in anything other than a producer/editor role I'd be shaking in my boots right about now.
I'm interested in how this is better / worse that runway
I'm very keenly interested on what their underlying tech/model is — I couldn't find any detailed information.
What do I have to get off the waitlist! Dying to check this out.
Pika is slang for pussy here in Iceland, just an fyi
This isn't really on topic for HN until there's something more than a waitlist for people to try out.
Otherwise the thread fills with comments about the waitlist and/or how there isn't enough information yet.