A Legit Sora Competitor is Here + Testing AI Actor Emotions
HTML-код
- Опубликовано: 19 июн 2024
- In this episode of AI Film News, we explore how the latest AI actors are becoming incredibly expressive and realistic, with cutting-edge tools from Adobe and Synthesia leading the charge.
Find all of the links from the video here: curiousrefuge.com/blog/ai-fil...
Time Code:
00:00:00 Intro
00:01:30 Adobe's Video Giga Gan tool and its impressive upscaling results.
00:03:06 Synthesia’s expressive AI avatars and their capabilities.
00:04:23 Potential applications of AI in video production and virtual environments.
00:05:38 New AI-driven video and image editing technologies.
00:07:31 Updates on various AI video tools and their applications.
00:09:34 Innovative AI video editing tools and their impact on the industry.
00:11:06 Competitive AI video generation tools and performance comparisons.
00:13:56 AI-driven filmmaking insights, events, and competitions.
00:16:33 Final thoughts on AI in filmmaking and community engagement.
00:19:16 Outro
Helpful Links:
Adobe Video Upscale details and issues: www.theverge.com/2024/4/24/24...
Project Homepage:videogigagan.github.io/
Synthesia AI’s New Avatars: www.synthesia.io/avatars
Reid Hoffman Cloned Himself: • Reid Hoffman meets his...
Scaling NeRFs: research.nvidia.com/labs/toro...
Text-based image editing features: x.com/dreamingtulpa/status/17...
Rendernet AI: x.com/Mr_AllenT/status/178533...
Interactive 3D Generation: interactive-3d.github.io/
SORA Competitor: / sora_competitor_shengs...
Shy Kids Breakdown: x.com/bilawalsidhu/status/178...
Haiper’s New Explore Page: x.com/HaiperGenAI/status/1782... 
Развлечения
Always appreciate your news updates.
Thank you so much!
Pumped for the AI Film Trailer competition! I'm thinking of giving it a shot! :D
Please do!!
Thanks for all the great info & updates, all around. Haiper is indeed pretty cool, and so many of these other tools, too.
Curious question: I'm wondering about the pronunciation of "Vidu". I've heard it pronounced differently by different people.
It's a VIDeo generator. Wouldn't it be pronounced as Vid-u (short i, short u), as opposed to vi-du (long i, short u). It seems pretty apparent the root word they're using for the name of their product is video (short i).
(I've just always had this odd thing about wanting to pronounce names correctly, as people very often have pronounced my last name wrong. Just a quirk, I guess.)
Good point, we'll confirm and get back to ya :)
Is the Thumbnail a "May the 4th Be With You" reference? If so, kudos!
Haha it is!
good stuff !
Thanks for watching!
So there's still hope for Keanu Reeves.
there is always hope for Keanu!
Why do tech enthusiasts exaggerate their titles. AI actors are not here. AI cartoons maybe, AI anime, possibly. But AI is very far away from achieving the nuances that would warrant a business actually using these tools to make a serious feature film.
It's annoying how tech enthusiasts exaggerate the reality of a tech they believe in. Similar to bitcoin and NFT's. Tech enthusiasts exaggerated the existence of the tech for years, the value of the tech for years, and it all ended up being inflated info, lies, and scams. I swear, just because something would be cool if it ACTUALLY EXISTED doesn't mean we should pretend it does. Sheesh people. ☠🤡☠
You clearly have no idea just how fast things have been moving in both LLM's and the generative-art world.
This is nothing like Bitcoin or NFT's, nowhere close.
I am going to assume from what I have seen this is either a sh*t post or you actually want answers. I will therefore assume it is the latter and while annoyed, you might actually want clarity on the subject. Therefore, the following "essay" will be a response to the inquiry. I have no intention of dropping a multi-page reply and I will do my best to summarize. However, I am currently under the impression that if I do summarize there is a chance the response may not be fully understood so if I feel it is necessary, I will go out of my way for context.
That said:
Lets get something out of the way right from the start. AI is an exaggeration. These machines are not "Artificially Intelligent". The definition for clarity is "the theory and development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages." Most programs in the past year and a half to two years are pattern reignition programs that use production logic to "generate" responses or results. Now, you could argue that this is the machine version of "thinking" and there is a possibility this may be factual, but that is something that is still undetermined. The goal in our minds is to achieve a thought process akin to a person sitting down and thinking in a human way. To some extent this just is not possible but on the other hand there is a lot of theoretical data that highly suggests it can get very very close.
And in a way we can establish that a part of this definition has been achieved. "AI" can switch between languages, we have visual perception modals, and it can preform a lot of tasks that tend to require human input. However, while individually a lot of these have been proven to be possible, we do not have a singular "AI" able to do everything mentioned. LLMS are great except when being given tasks as it is limited to only preforming tasks on a computer. Robotics has come very far and we are starting to test integration of LLMs into robots, and that will be the next big leap towards a true "AI".
So, why are we calling it AI anyway? Because the technology is so borderline close it will eventually be called AI so why not start now. Why come up with a whole different term just to call it something else down the line? And to be frank, we do not have a term to define something that is Borderline AI but not AI. It would have to be invented. Which would be a task on its own getting people to agree to a term. So, shortcut the arguments and spending time on it, call it AI because it will be eventually anyway and move on.
With that established, lets also take a look at WHO "Tech Enthusiasts" are. Because they are not the people behind the closed doors developing this tech. An Enthusiast in general is basically a supper fan. They love something more than the average consumer would and are thus willing to put more time into research, using, learning, etc, than an average consumer. With that out of the way, the simple answer is "welcome to the modern world of marketing". The complicated answer is pay attention to where you see these sort of claims coming from. Our Media Market is pulled by who can attract the most attention. It does not care "how" that attention is attracted. You simply want the attention on you. Enter in the world of bogus video titles and making claims with nothing to back them up. And the art of throwing out complicated words and names because it sounds cool or is trending. Fact no longer matters here, attention matters.
Result: We get what you have complained about. For every piece of factual data on the internet there are hundreds of twisted, altered, deformed, data about the same thing that are lies, overexaggerated, etc. Why are we having a freak-out of people opposing AI thinking it will end humanity? I will tell you right now it is not because those people did any research or understand how technology works. They got online, saw an article, saw an opinionated twisted concept of what could happen, got convinced by the same person it will happen = mental crisis. This is just how marketing works in our day and age and if we want it to change it will, but it will take a generation or two and the only way it happens is if the common concern of the masses is to care more about facts and data than being convinced something is something. The need to "sell" prays on manipulation and convincing your target audience you are right or have the right answer. It doesn't care if you are being given factual data.
@@JustMe-ty2rp you had to have dropped out of grade school before taking English right? It would behoove you to take an elective or two on reading comprehension. In the interim allow me to illuminate your dented response; it did not address the critical point of his comment, woefully exaggerated clickbait titles. It's complicated, I know.
I chatted with the CTO of CAA yesterday about their AI actor scanning program. AI Actors are here and they are very much being used in professional projects. The user-facing tools demoed in this video are just the surface of what is possible.
"ai can do cartoons, but not a serious feature film" uh, cartoons are one of the biggest businesses in the world.
Thumbs up for the mustache👍
thanks so much!
How about ‘Story Diffusion’?
It’s an open source AI tool, so I’d love to learn more about it.
We'll check it out and get back to you1
Thanks
Our pleasure!
can't find links to ai films 😭
Check out our website competition page and we'll link to it :)
Hello, links,to,the films?
Check out our competition page on our website and you can see them all :)
If the creators could add a cinematographer detail box this could be very cool, seems the lighting isn’t consistent from one shot to the next in the sequences and that mean AI is lighting the scene and we have very little control of that.
Good point
Thanks, but is there an AI that can apply the change of clothes to the video and not to the photo?
You can try Pika Labs as they have video inpainting
@@curiousrefuge Thank you
@@curiousrefuge Do you have tutorial for Pika labs video inpainting? I tried it but result wasn’t good at all.
where are the links to the films???
On our website you can find the competition page and you can see it :)
I cant find it. 😫 its okay. Someone posted one of them somewhere.
Mich interessiert auch, wie es nicht nur in Englisch funktioniert, sondern auch, ob andere Sprachen vernünftig umgesetzt werden können, beispielsweise mit deutschem Prompt und in deutscher Sprache. Vielleicht kann dieser Aspekt in Analysevideos einfließen als Anmerkung. (Avatar AI Features)
that's a good question! We haven't done any german testing but we can try it out and get back to you :)
300:1 ? 300 attempts for a good take? Does this take 12 days creating variations 8 hours a day to use a take? Am I crazy or did I get it wrong? What does this tell us about the workflow of this tool if you are going to create a short or marketing material using only sora?
Remember, that's 300 generations of a specific style/look on someone's personal computer and they artistically chose 1 even if there were plenty of good ones. This is the worst this technology (still basically beta btw) will be - so in the near future that ratio will go way, way down. All of this compares to legacy production methods that cost way more time/money.
I already love the AI generated 50s Panavision movie trailer popping up all over the place, cant wait for feature length fan films...
We can't wait too !
VideoGigaGAn is public avaiable?
Not yet, we'll get back to you!
Your discord link?
You can find it on our website! It's a free community :)
The problem for me with the AI images and video is that it looks more like a persons memory of a scene than a real moment.
True, there is a halo effect that is hard to get rid of. However, as these tools advance they may be easier and easier to solve.
A billion dollars for an avatar company that will be replaced in a few days.
The industry will certainly be disrupted !
300 to 1 for the sora shots the shy kids used suggests the current sora examples are very cherry picked, I wonder how good the average shot actually is with it
It probably took quite a few iterations, but this technology will improve drastically over the next year.
The Ai figures look like over actors
It's not quite there yet but it will be soon :)
The avatars have the wrong voices, they don't sound like they look, the young woman has a middle-aged woman's voice and there's no ums and arrs, that is why they look uncanny
As the tech gets better, so too will these demos :)
Why do young people dye their hair to look old? I don't understand.
That's legit my hair. I've been going grey since I was a teenager. haha
@@curiousrefuge Whoa really? Oh okay then. Sorry about that. I have seen some other young people dye their hair silver/gray and it just makes no sense to me. It looked really well done and I thought it was a very elaborate grooming dye, so at least in that way your legit hair colour had me fooled you had it done up by a professional stylist.
300 unusable shots to get 1 usable shot using Sora? That’s really not very good.
True, it's quite a workflow. But compared to some legacy workflows like thousands of dollars, time, etc. It's an interesting comparison
Far from Sora, come on. This is not a "legit Sora competitor". The quality of the images is nowhere near Sora.
Agree, It's mind boggling to me that people even compare the two...
We understand, but we think there are quite a few shots that could be very comparable. We agree to disagree!