OpenAI’s Sora model is making waves in the world of artificial intelligence, with creators and tech enthusiasts alike getting their hands on it to generate stunning videos from simple text prompts. In a recent video by tech reviewer Marquez Brownley, also known as MKBHD, we get a glimpse into the capabilities of this impressive model.
The video showcases nine brand new examples of Sora in action, from a red cube being printed by a 3D printer to a product reviewer surrounded by gadgets in a well-lit environment. While Sora’s generated videos still struggle with some aspects like physics and branded items, the lighting and shadows are remarkably realistic.
The creators of Sora are fine-tuning the model to ensure safety and prevent its misuse for propaganda or political purposes. While the model is currently invitation-only, it is expected to be available to the general public by the end of the year, with plans for a limited number of prompts per day.
Overall, Sora’s ability to generate lifelike videos is a significant step towards achieving artificial general intelligence. It’s both exciting and a bit scary to see how far AI technology has come, and many are eagerly awaiting the chance to explore Sora’s capabilities firsthand. So, stay tuned for more updates on this groundbreaking AI innovation!
Watch the video by Vanis Jakub
Video Transcript
Good morning and welcome everybody to my channel my name is yakub vanish and we’re going to be talking about open AI Sora model so this is my third video on the topic because as you can see I’m very fascinated by it uh today we’re going to be talking about nine new
Videos that surfaced uh recently because the creators more and more creators are getting you know their hands on this and also one of my favorite creators Marquez brownley MKBHD uh got his hands on this and he tried three prompts and three videos this the first one that we’re looking at
Is a Red Cube being printed by a 3D printer um and yeah as you can see physics sometimes don’t match in this one but it’s still very impressive the next one I think is by far the most impressive uh that’s a young reviewer product Tech reviewer in a well lit
Environment surrounded by gadgets holding a cinema camera so as you can see it’s getting better and better this is almost in the indistinguishable from reality and there are some learnings that MKBHD was sharing he says that it can struggle sometimes with physics especially with legs or or fingers and
Walking uh sometimes the Sora generated videos you know they struggle with walking because the legs cross over and merge into each other and that that gives you that gives it away basically and also it struggles with branded items they never quite match up to the real
Life things like cars cameras ET ET they are not identifiable as a specific model on the other hand what he was praising is the lighting uh especially in that video of that product reviewer it’s almost as if there is a large soft key light just out of frame to the left and
Across the board lighting and shadows match all over the place so it’s it’s really really nice and impressive uh he also spent some time with the with the developers uh from Sora from open Ai and they shared uh some bits and pieces they were saying that they’re not just ready
Yet to to kind of release this as a finished product so they are planning to do it in the coming months um and right now it’s an invitation only thing uh and as you can see MKBHD just got you know three prompts and what is surprising is
It it took quite a while to generate them he says that you can go out you know to Starbucks get a coffee come back and it will still be working so we’re most likely talking about hours to generate each prompt and most likely the model that they’re going to enhance or
Utilize for this is that they will allow you to you know use maybe one or up to three prompts a day or something like that and that’s it I don’t know if they’re ever going to make it as part of the CH GPT as as do 3 for example
Currently is or if it’s going to be totally separate but currently they are fine-tuning it um and especially they are really uh looking into safety aspects right right because this can be easily used for some propaganda political issues especially you know there are so so many big elections
Happening this year around the world right in the US currently also in Russia so so all of these things they are taken into account and I think it’s going to take up until almost the end of the year until general public is going to be able
To to test it out and I think they will likely start with uh you know one prompt per day maybe for for CH gp4 users just to do like get the taste a demo taste you know for free but in the in the future it might be you know you will pay
Su you go per each clip currently it can generate 60c Clips as you can see this is an alien walking through New York City and it’s just quite impressive of course yeah it gives it away but it almost looks like a guy with a real mask walking through New York City another
Interesting fact that they mentioned in the interview was that uh this model Sora is allowing them to to reach AGI faster because currently the llms the famously the CH GPT 4.0 you know that one uh already has a huge understanding of the world um di also helps it to
Understand in terms of pictures you know what’s what’s going on but the videos this is how each and every human kind of views the world right like in movement in motion um and this this is adding a huge layer of understanding to these models to to artificial general
Intelligence to be able to actually uh to be able to actually emerge basically so yeah that’s uh I don’t know if it’s scary or exciting to be honest uh but but I’m looking forward to get my hands on this and as always this has been yub vanish thanks for all the follows any
Comments I’ll definitely reply to them and I’ll keep coming back to you with more stuff like this so thanks a and cheers everybody
Video “OpenAI’s Sora – 9 Brand New Examples!” was uploaded on 03/13/2024 to Youtube Channel Vanis Jakub