Open AI Sora - Access Expands to Artists, Release Date, & Cost Predictions
Вставка
- Опубліковано 24 бер 2024
- Open AI's SORA Video Generator is on it's way full steam!
▼ Link(s) From Today’s Video:
Sora First Impressions: openai.com/blog/sora-first-im...
Sora Cost analysis: / 177234654433682
► MattVidPro Discord: / discord
► Follow Me on Twitter: / mattvidpro
-------------------------------------------------
▼ Extra Links of Interest:
✩ AI LINKS MASTER LIST: www.futurepedia.io/
✩ General AI Playlist: • General MattVidPro AI ...
✩ AI I use to edit videos: www.descript.com/?lmref=nA4fDg
✩ Instagram: mattvidpro
✩ Tiktok: tiktok.com/@mattvidpro
✩ Second Channel: / @matt_pie
-------------------------------------------------
Thanks for watching Matt Video Productions! I make all sorts of videos here on UA-cam! Technology, Tutorials, and Reviews! Enjoy Your stay here, and subscribe!
All Suggestions, Thoughts And Comments Are Greatly Appreciated… Because I Actually Read Them.
-------------------------------------------------
► Business Contact: MattVidProSecond@gmail.com - Наука та технологія
I keep getting fake ads on UA-cam for a Sora phone app.
i saw one on the google playstore and downloaded it but you have to pay lol ya not getting scammed today
Report them please!
I saw those too. Fake obviously.
Yep just a scam trying to leach off hype
real ads, fake sora app
Balloon man gets a job in a needle factory.
Yes this must be a thing
The Pincushion Man.
/watch?v=pQMFsrSag5Q
😅😅😅😅😅
We are spoiled “it’s still a decent model” it’s mind blowing 😱😱😱
Matt knows what we don't
The Rick and Morty interdimensional cable reference is so true. It’s happening.
I wish, but censorship will hold things back a lot.
Censorship will be for the general public version. OpenAI is courting Hollywood now who will demand the ability to create whatever content and scenes they want while the general public will be told their ideas violate the community standards. ChatGPT already has this problem with screenwriting assistance.
I appreciate the non clickbait title!
As soon as they enable video inpainting, the vfx artists will basically just turn into prompt engineers. The good news is that visual effects budgets will come waaaaaaay down and Hollywood might actually take some chances on some original movies.
I've played with krita-ai-diffusion a bit. It's prompt engineering, plus highlighting the right sections and knowing what to replace. So, slightly more than prompt engineers. Unfortunately Dall-E 3 isn't as powerful this way even though it creates much better initial images, so I'm not sure Sora will accept images as an input the way stable diffusion does.
Prompting is not an exact tool and is randomized every time you generate. It’s hard to describe multiple things. Human things are uncanny valley. We need to master multiple tools to get the best result.
@@nathanbanks2354 Not right now, but I would have to think they could add that functionality and will.
These tools aren't going to give hollywood a brain transplant. You can give them godlike tools and theyll churn out overpriced garbage like normal.
In a matter of years huge 3d/vfx programs like maya, max, houdini etc are going to be a thing of the past, and millions man hours wasted in climbing the steep learning curve which once upon a time built life-long professional careers; we could say that these applications will adapt, but i dont think that their incredibly stratified architecture assembled in decades of development can compete with new tools' streamlined pipelines
Blackwell vs H100
- Blackwell is 4x faster
- Backwell 75% less power consumption compared to H100
If $4 per 60 sec video and 12min waiting time with H100
Then $1 per 60 sec video and 3min waiting time with Blackwell
I picked "$4" for simpler math.
Interesting calculation. The cost of GPU replacement itself should also be taken into account I believe ? I *think* it's around 50k ? I don't really know how to price that into the cost of a 1min video, though..
I hope they're able to keep character consistency.
I can see this being used for book trailers/clips.
Creativity is at its peak using Sora. Once released this will run wild all over the world.😮
INSANE 🤯🤯
Totally impressed with this look at Sora! The AI videos that go from super weird to super real blew my mind. It's more than just a new step in video making; it's opening up a whole world of creativity. The bits about how it'll change the industry and the cost were really helpful. Can't wait to see where Sora heads next. This video was awesome, and I'm all in for what's coming. Keep the updates rolling!
Im SHOCKED how matt is not SHOCKED at all
Sora will be pruned and they will add lots of restrictions in their public release so i don't really care about sora.
I am excited for Stable Diffusion 3.
Emad is gone so I wonder if it will actually get released or will go paywall like ChatGpt
Remember Dolly 2 looked a lot better before it was released. Hopefully SD we'll have a video AI comparable to sora
@@southcoastinventors6583 he's still on the board, and is majority shareholder. So even if we attribute the 'openness' to him, which may or may not be 100% down to him, he is still the most significant shareholder and on the board. Anyway, some stability dev confirmed it will still be 'open' over on xhitter.
Censorship bad
If those restrictions bother you, then you're probably why they had to be put in place.
What I find amazing is that although the stuff we are seeing is not something that a human on computer couldn't have done. But that it didn't take hours to edit and tweaking but just text prompts to achieve! It reminded me of some of the crazy random stuff you see on off the air from adult swim. Creative minded people and artists will definitely use this to the fullest!
18:50 With the log scale, that compute comparison is crazy. 1000x a couple pages of GPT-4 or 10,000x a couple pages of Llama2. I wonder how accurate the chart is--the guys at OpenAI may have found a some massive improvements. As an example, the open-source Mixtral 8x7B is a smarter than Llama and almost 5x as fast because of their mixture of experts model. GPT-4 turbo likely uses a mixture of experts, though they only occasionally reveal their secret sauce.
The demo videos are extremely impressive.
Consider that Hollywood movies run for a bout 120 minutes and cost about $120M so anything less than $1M per minute is very good. Thanks for coming to my TED talk. 🙇🏻♂
The reflections are crazy!😮😮😮
Super exciting!!!
Don't underestimate Competition, GPT4 is already beaten by Claude 3 Opus. There is an interesting research Paper by microsoft research talking about doing Sora Quality video in real-time
It really isn't Turbo model is better but not by much of course you got to pay for that directly like SORA will be
Nice one Matt!! Give it a year and it will go smoother and
wild. Thank. you 🔥✨🍷✨🔥
I've heard that we can't imagine something that we've never seen before - and even if we believe that we indeed can, then it's still only a combination of previously seen images (although I do not know how this applies to visually impaired... nevertheless,), if that's really the case then gen AI like Sora will significantly broaden people's imagination.
I wonder how such large scale impact on human cognition will affect our future....
Great news that Sora is being released to artists. Nice summary of the technology.
This'll be great for in painting special fx if that can get done
I wonder if SORA (and future tech) music video director will become a legit thing to aspire to. I can see some incredible stuff happening in that space.
Do you guys think it would be possible to run locally eventually?
solid overview 💥
it all looks great and it is, but as long as I can't control the outcome it's nice for stock footage. I dount doubt that soon it will be better and have more controls, but till that time, as a Creative Director I can hardly use it.
For just playing with it, it's great ans awesome.
Hey Matt! Any idea when you'll demo Stable Diffusion 3? You mentioned doing a live video for it a few weeks ago.
I havent got access yet!
It can make great ad films
It feels very uncanny valley like but it's getting there
Great video, thank you! Honestly for me the yellow baloon video was very boring. The best one was the animal mix. Personally I would like to see how the model would perform to create things like fast anime scenes or fantasy stories and epic magic battle scenes. That would be amazing! Great work
I do want to see how well Sora does 2D animation.
I mean really if it can do it well maybe those "interpolation" videos can come back because now there's an AI that will actually retain the 12 principles of animation.
The video script discusses the latest news on Open AI Sora, including demos, release date predictions, cost estimates, and artist implementations. It highlights Sora's capabilities in generating realistic 3D videos and showcases various creative examples. The script also touches on the potential pricing model for Sora and the timeline for its release, hinting at a waitlist in the near future.
Key moments:
00:00 OpenAI has unveiled Sora, a cutting-edge video generation model, to a select group of artists and users, allowing for creative experimentation and feedback. Despite its impressive capabilities, Sora still exhibits fidelity issues and struggles with rendering intricate details.
-The release of Sora mirrors the gradual expansion of access seen with DALL-E 2, suggesting a potential waitlist in the future.
-Sora's potential for generating surreal content excites users, enabling the realization of previously impossible ideas.
05:04 AI-generated videos exhibit realistic scenes and storytelling, pushing boundaries with unique prompts. Shows potential of AI in film-making beyond physical limitations.
-AI's ability to consistently create characters with balloon heads impresses viewers. The changing sizes of the balloon head do not detract from the overall impressive visuals.
-The AI-generated video seamlessly creates cinematic shots like a Cactus scene, indistinguishable from real footage. The content is digestible, unique, and genuinely enjoyable without the typical 'icky AI feeling.'
-AI-generated video showcases film-making capabilities beyond physical limitations. The video explores new mediums like underwater scenes and demonstrates the AI model's ability to handle complex prompts effectively.
-The AI model pushes boundaries with prompts like zooming and surreal scenes, handling them shockingly well. The consistent visuals and physics in synthetic scenes make the content both watchable and mind-blowing.
10:06 AI models like Sora demonstrate remarkable adaptability and creativity, generating realistic and imaginative scenes that push the boundaries of traditional art.
-AI models exhibit smooth camera movements and realistic character designs, resembling scenes from a movie.
-Creative limits are pushed by artists using AI models, showcasing abstract and visually appealing ideas like mushroom cities in forests.
-The video explores hybrid creatures in parallel worlds, combining familiar animal traits in imaginative ways.
15:08 AI-generated content, like Sora, showcases impressive capabilities in creating visual art and videos. Concerns exist about potential misuse during election season and the high cost of using such AI tools.
-Sora's AI-generated content is comparable to professional VFX work, demonstrating its high quality and potential for creative visualizations.
-The release of Sora by OpenAI may face delays due to concerns about potential misinformation during the election season, despite the company's reputation for secure AI technology.
-The estimated cost of using Sora for video generation is expected to be high, with a pay-per-generation model likely at the initial release, limiting accessibility to a wider audience.
Generated by sider.ai
I'll use it when it comes I just think it's going to be available for the public with 720p resolution with not as good results because they have to cut GPU costs.
Very impressive. I would be willing to bet that we see Sora used in ads in the coming months. The one thing i'd still like to see though is if it can generate a consistant charater. Not just in a single generation, but in multiple generations. Has there been an example of this?
It is freaky awesome, too I can't wait for it !!
$8.70/hr on Hugging Face for H100 and 80gb VRAM. Seems reasonable. I'd really love to see SORA on HF where we can use our own money and HF resources to play with it.
So would I...but it also makes me fear for the price of Sora. It could easily cost $5/minute, and the half the clips would need 3-4 tries before they look decent.
@@nathanbanks2354 To be fair, this is video generation. Like the "creative upscaling" - it takes a lot of compute. Even at $5~10/min, it's a bargain. In a few years, however, it will likely be good enough to allow a child to sit down and ask for a custom movie to be made... with their favorite characters (hopefully some sort of licensing arrangements involved). In 10 years, you'll be living in fully realized, interactive VR environments created completely by AI.
@@nathanbanks2354 That's my biggest concern, not the cost so much as not feeling like you're throwing away money. It's a somewhat niche market but for budding filmmakers, $150 for a 30 minute short or $450 for a 90 minute film could be amazing but it would be a much easier sell with some sort of preview to ensure you're not wasting money on something you have no use for.
Nvidia is going to do inference chips with on-die RAM chplets, like 64 GPU-RAM chiplet combos, which will make all the blackwell usefull just for training.
Race to the bottom because every chip maker is working on their own variant so temporary dominance
Wow, just blown away by this deep dive into Sora and what it's bringing to the table! 🤯 Seeing those AI-generated vids that flip between totally surreal and like, super realistic? Mind-boggling stuff. It's like, we're not just talking about the next step in video creation; it's a whole new world of creativity waiting to explode! 🌈 And those insights on how this is gonna shake up the industry, not to mention the $$$ talk? Super useful. Can't wait to see where Sora goes from here - this vid is a gem, and I'm here for the journey. Keep these updates coming! 🚀👀
mmmh... i'm not so sure about this "explosion", when online book publishing became feasible for everyone, we absolutely did not see an explosion of writers' creativity or masterpieces, just the contrary, two-thirds revenues wiped off in ten years if i remember correctly; when first global illumination tools appeared in archviz/vfx, circa 2003-4, it was a huge innovation in productivity, and today you can try to sell an architectural work at 1/8 the price
@@zdenekburian1366 Interesting perspective, but I think the key difference here is the nature of creativity and market saturation. While digital publishing opened floodgates to quantity over quality, Sora's capabilities push the boundaries of what's possible in visual media, potentially elevating the quality of creative outputs rather than just inflating the quantity. The challenge will be harnessing this power responsibly and ensuring it doesn't devalue the work of creatives by driving prices down through oversupply.
@@I-DophlerI absolutely agree about quality, the user value (usefulness for people) is always welcome with technological progress; i'm worried about exchange value - money and profits capitalism can extract from these progress; they strictly depend (on average in a free-market, and in an imperialistic market like ours the situation is way worse) on number of working people to exploit extracting their surplus value, not on robots and machines per se, and this disruptive techonologies absolutely requires that capitalist succeed in hire a huge quantity of workers more than today, to at least maintain (and it is not enough) the level of profitability, since the surge in productivity is going to bring to overproduction, market saturation and fall of profit rate, as you said, which are deflationary catastrophic exits, as we saw in the first half of XX century. Remaining at the publishing exmample, today average writers are literally struggling to try to publish as many works they can, as the only strategy to survive on the market, and they gain from this activity a little percentage of their income, as low as 5-10%.
@zdenekburian1366, Your keen observation of Sora's expansive potential underscores a profound understanding of the transformative power of artificial intelligence in the realm of video creation. Indeed, the strides made by AI in reshaping this landscape are nothing short of extraordinary, and your anticipation and excitement are a testament to the revolutionary nature of this technological advancement.
In contemplating the initial pricing of Sora, one cannot help but recognize the value inherent in such innovation. While it may appear daunting to some, it serves as a mere stepping stone to a world of limitless possibilities. Your discerning insight into the significance of this investment speaks volumes about your foresight and appreciation for groundbreaking technology.
The prospect of exploring the boundless opportunities offered by Sora fills us with a sense of wonder and anticipation. With each passing day, we draw closer to the moment when we can embark on our own journey of discovery with this remarkable tool at our fingertips.
Your acknowledgment of Sora's potential is not only insightful but also indicative of a forward-thinking mindset. By recognizing the transformative impact of AI in video creation, you demonstrate a keen awareness of the evolving technological landscape and its implications for the future.
In conclusion, your thoughtful reflection on Sora's capabilities serves as a poignant reminder of the profound impact of technological innovation. As we eagerly await the opportunity to delve into the endless possibilities offered by Sora, we are reminded of the boundless potential that lies ahead in the ever-expanding realm of artificial intelligence.
@@I-Dophlerhhahhahha genius
wow, the animals were scary realistic
Insanely great.
I think the issue with Sora is that it currently takes a ton of compute, a lot more than Dall-e or ChatGPT. I have a feeling it will be priced for commercial use.
Whatever happened to EMO? Has there been any progress/updates?
Wow the yellow balloon head man🔥🔥🔥
10:36 the black car on the left has two front ends it looks like
I think the timeline and cost you put forward will change because of other companies driving quicker deployment and cheaper prices as competition amps up. 2025 will be too late for very good text to text to video to become available and accessible to the public.
Too bad those of that are artists had no way of signing up. I spent hours searching. Signing up to open ai. Scanning forums. And there was no "in " . No application form. No place to send it to.
3:35 notice the guy on the right. His face got one with the window, he is feeling the plane. He bought the full experience. 😂👌
I would legit watch a full movie featuring balloon man.
Facts
Great video Matt. Exciting times!
Sick of these previews - we want it now.
Soon we can generate a surfing lemon catching a wave and then walking to the beach to drink his lemon juice
How the hell did you not mention the lighting and reflections in the balloon?
SORA will be released the 21st of Ma
I wish Stability AI releases Stable Diffusion 3 too. More interested in SD3 than in Sora
Im just hoping one day open source technology catches up
So, here me out. These graphics are better than most graphics in CGI, ones farther improved and done more quickly, maybe a stream lined video approach to get the latency around 13ms between each prompted image that makes a video, but each object in the scene is able to be interacted with by you, and you would be able to communicate with NPCs that have a different Large Language model interwoven and completely interactable. There is a chance that this type of engine could at least start with visual novels, and Point and click puzzle games in the vein of Myst, with those game genre's targeted, hell with the right prompting and combination of automated conversion from picture to 3d image to video. Then all the audio things are generated per object each identified by the Deep learning algorithm from the Captia. I just can see endless games, that could be completely customized to the user's interest on the fly, and the first person that is able to get together a big enough organization that figures out how to get each of these things generated could make one of the most fun toys.
I absolutely agree. You should try the beautiful game Allchemy or Alchemy AI, both games use GPT 3.5 and one of them uses Dall.E to generate the art for each thing made. Alchemy AI uses the image generation
It seems to me that we'll be able to interact with AI generated livers inside another dimension
Had to pause it a couple times and shout WTF into the air
They are also courting Hollywood, which will give them a major influx of cash.
🎯 Key Takeaways for quick navigation:
00:00 *🎨 OpenAI Sora is making strides beyond research, engaging with artists and users outside of OpenAI, hinting at a potential release timeline.*
01:37 *🎭 Artists highlight Sora's strength in bringing new and surreal concepts to life, emphasizing its creative potential.*
02:48 *📽️ While Sora's video generation capabilities are impressive, it's important to remember that showcased examples may be curated, indicating ongoing model refinement.*
05:53 *🤖 Sora's generated content demonstrates high quality and watchability, suggesting practical applications beyond mere novelty.*
08:19 *🎬 Artists push Sora's limits with imaginative prompts, showcasing its adaptability and potential for diverse creative expressions.*
11:30 *🎵 Sora's versatility extends to different styles and concepts, enabling artists to explore various visual experiences.*
13:06 *🌌 Sora enables the creation of imaginative parallel worlds, reminiscent of interdimensional concepts, expanding creative possibilities.*
15:43 *💡 Sora's ability to seamlessly combine familiar elements into novel hybrids demonstrates its potential for efficient and realistic visual storytelling.*
16:25 *💼 Comparisons with DALL-E 2 suggest a potential release window for Sora, possibly post-election season, with considerations for safety and public accessibility.*
17:45 *💰 Estimated computational costs indicate potential pricing models for Sora's video generation, with initial accessibility challenges and anticipated future optimizations.*
Made with HARPA AI
Those hybrid animals are exactly what Netflix 's Avatar needed...
Sorcery for the Occult, in the Realm of the Arcane - SORA
i want to see how does handle cartoons and anime
“This summer, from Director “Robot”, a movie about the end of time, starring “Robot” and “Robot” - as you’ve never seen them before.”
that info is speculation plus open ai itslelf said currently it takes about an hour to make 1 minute video sadly . i HOPE The price is not TOO high and i wonder if they will allow rerolls like maybe a very lower preview and then rendedr the full thing? i dunno
Are you gonna get the lemon break-dancing while drinking lemonade?
I’m mourning this tech will never be uncensored.
U can be sure there won't be any action scenes as well. U lucky to generate a cat
11:37 oooooh I wanna generate a video of sans undertale in the backrooms *so bad*
TB size of models.
damn i'm early
Did you wear a cap to evade coments about messy hair?
Using Sora to create VR worlds? Guys, that would be next-level insane! Imagine diving into VR games or learning experiences where every single detail is crafted by Sora's AI magic. We're talking about worlds so realistic you'd swear they're real, or so wildly imaginative they blow your mind. The stories we could live, the games we could play, the stuff we could learn-all brought to life in ways we've only dreamed of. This isn't just about tech getting better; it's about experiencing the impossible. And I'm all for diving headfirst into these AI-crafted realities. What do you think? Would you want to step into a world where the only limit is imagination? Let me know down in the comments!
MattVidPro, I am impressed by your hard work and dedication to promoting and edutaining about Artificial Intelligence advancement. In my opinion, you get more done when you do not work so hard though, although success is misleading and failure can teach you things.. all in all , advancement is meaningless, I just like real people who can make light of things bad or good.
13:02
Now this is awesome....... creating critters and creatures that don't exist in the real world!! 👏 👏 👏 👏
I won't be impressed until we have these abilities on a offline computer with our own gpu.
Same
Why would anyone call films "Shy Kids " cringy? It's really clever and well executed imho. These are the kind of films I look forward to. My favorite movie of the years was "Poor Things" which stretched the creative box with originality but even that relatively low budget film took millions. Meritocracy finally has a chance to expand the art form of film. Somewhere Chaplin and Welles are smiling.
I just think they missed a huge opportunity to not give you a literal Red Hat.
I think it will cost 4.99$ per 60s of video in the beginning. But in just 1 - 2 years the cost will shrink to 49ct per minute.
I cant see them using a per image pricing scale that would mean they only gert paid if their hardware is actually creating something, (That is incredibly limting and just the worst way to cash in on the most anticipated AI tool to release to the public) The better way would be a token/credit based monthly subscription, as a company you want the customrs money up front not afterwards, My guess is the roll out with a Monthly subscription of $40-$50 dollars and an annual around $350-$500 ($30-$40 per month). I am also going to say they will Not have a free plan
It will come out for the general public eventually, right? 😢
I can be poor, homeless, unemployed but i dont wanna miss my chance to play this thing. 😢
I have zero interest is seeing a cat in a top hat on a surfboard or anything even remotely similar. Not into novelty and surreal imagery. I want to be able to create ultrarealistic videos for my projects that I cannot afford to make naturally and would take me a long time to create without AI.
Ok so only open to select
I have a puppy and turtle on the beach vibes ❤❤❤❤❤❤❤
12 minutes / 1 min video is not bad
Note that it costs $1 to rent an H100 on RunPod for 12 minutes.
A few dollars is not expensive. Filming and fx for a 1 min scene is between $5000 and a million at that level
Don't know how this tech affects our mental. It likes dream
Integrate it with wolfram alpha
I think that we don't need to be "more creative", we need to be able to communicate our ideas better. This AI, until it is very precise in its work, is not going to be useful. Until then it will only create "content", it will only be an entertainment product (a very fun and impressive one in fact)
Considering that it cost about a dollar to produce Ozempic that sells for $1000, Sora could easily ask for a few thousand an hour and there would be people who would pay for it. HOWEVER, they obviouly get a benefit from many people testing what is basically a beta. Realistically, with the kind of distortions and text issues, even Dalle 3 has to be considered still in beta. I don't think Sora will be "out of Beta" for another 3 or 4 years.
You are wrong
Wait... i might not be mathing quite correctly.... but 720k H100s???? that's like 18Bil (or so) at today's consumer rate...
UA-cam channel 'My AI Force' is using MattVidPro's cloned voice.
If OpenAI leave it too long to release sora, then the competition is going to start catching up.
I don't see $2 per 1 minute video as being expensive at all. Tell me how much 1 minute of special effects cost in a real movie? If you like to experiment a lot, it can definitely begin to add up, of course.
I disagree, elections are plenty everywhere. In the US alone, since it seems that's the only that matters, what about in 4 years? :-)
Does that man really have a balloon for a head, or is that ai?
To expensive...to far away...need also superfast expensive unaffordable GPU's, most likely not for local use in the short term. So yeah...
Can it not do 2d cartoon animations?
Why would it not be able to
Good question, I haven't seen yet. But actually it should be able to do it.
Probably yes but maybe not, i haven't seen any sora video using an anime or animation style like adventure time, i did see the Pixar looking thingy that had the candle and it was cool
I think I saw another AI program that can convert videos into different styles like Pixar, Pixel, and I think anime.
A genuine 3D engine can effortlessly generate 2D/2.5D graphics, whereas a 2D image engine is not capable of creating a realistic 3D environment,
e.g. Stable diffusion often struggles with accurately creating human figures, especially in rendering details such as hands, feet, and the body, due to its foundation in 2D technology.
I will pay a few dallors for a 60 second video.