I wish this AI tool never existed. Sora is dead?
547,183
Published 2024-03-28
EMO AI: humanaigc.github.io/emote-portrait-alive/
Clean up YOUR data with Incgoni at incogni.com/pricing enter promo code STAYINCOGNI and get 10% off our monthly and annual plans!
FOLLOW INCOGNI
X (Twitter): twitter.com/incogni_com
CHAPTERS:
00:00 Intro
00:35 EMO: Emote Portrait Alive
01:03 Examples
02:51 How does it work?
04:14 Dangers
04:45 How to become immune to it?
05:38 Advantages
Video made by Incogni channel.
#incogni #emo #sora
All Comments (21)
-
You look realistic.
-
So Mona Lisa speaks English, perfectly. And even when talking she still has that magical smile. Very impressive.
-
Really well put together video. Pleasantly surprised to see its a small channel 😊
-
Great video!😊The videos look super realistic
-
imagine a game where the story changes based on what you do, whatever you do, and you can do whatever and interact with everything even if its just a point and click, and everything is rendered as realistically as this
-
I remember maybe 5 years ago, mind was blown when paintings did a little head movement thing...all this ai stuff is already super nutty, and we havent even scratched the surface
-
Subscribed! I am so happy I found this channel.
-
Damn scary. Organised crime could have a field day.
-
This is impresive and terrifying at the same time. This technology could make you or someone you love "do" something you or that special one haven't done, with terrible implications for yourself, or that someone. And politics don't seem to understand how dangerous this can be without the correct legislation to protect human beings.
-
Very interesting video! Thank you for sharing this with us 😊 By the way, I believe that Chinese company's name is pronounced "Ah-lee Bah-Bah" - with the Ali Baba pronounced as two separate words. Great video! Thanks again 😊
-
Great video!
-
So we will, within the next 18 months or so, have models that generate emotive visual and audio output based on historical contextual sentiment. That means an AI that operates using more than a single prompt, but rather uses the history of experience with the interlocutor, and environment. In human terms, we call that a "personality". Couple that with some anticipated performance updates to the underlying software and it could be real-time using consumer-attainable hardware (though expensive enough to be beyond the reach of most, at first). I'm iffy on when the algorithms we use now (in open source) will advance to the point that the AI will have integrated short and long-term memory (for those that are unaware, AI models don't remember anything beyond their short context window). That's a fundamental change from what we have now (no, RAG is not the same thing as memory). But it's coming. It'll need to be more than just expanded context windows. It'll need to use experience to learn in real-time. That's all within 3 years (though I suspect it's here already behind closed doors). Then we get into some interesting philosophical (and eventually, legal) territory.
-
It's useless when you can't use it.
-
Hmm, when its available I would use it for my projection of my music performance. I have an AI that sings and compose, but the code doesnt support a visual, so the visual of my AI is pure offline Video-Edit, and with EMO I could improve the visuals.
-
For the Chinese video there is still uncanny eye-valley. But when the eyes are closed, man, that looks very compelling. It would seem the eyes are still the undiscovered country, so to speak, and are the last obstacle to continuous believability in video, across all types of scenes and camera angles. There's just that unnatural "dead"-ness that periodically creeps in even though the rest of the face is realistically expressive. Imagine if you were speaking IRL to someone and once in awhile, that blank/dead-ness crept into their eyes while in the midst of crying or laughing or otherwise exhibiting some intense emotion. It would genuinely creep you out.
-
You got a new sub! ❤❤
-
4:53 i have stopped posting photos & videos of myself YEARS now! No voicenotes sent either, YEARS. I only post what is in front of me, and most times that is not people.
-
The consistence between frames it's really impressive
-
There's some interesting potential here by combining it with the Suno AI thing Moist Critical was showing not ago. Could have the latest pop stars singing about a grilled cheese sandwich, or perhaps about constipation or whatever.
-
I just noticed The Prompt for the Sora video did not specify whether the Tokyo Street Scene was a daytime or nighttime scene. It only implied it through the mention of warm neon lights and signage. How specific does this have to be? Could you, in effect, give Sora a modified movie script and have it create a short film?