1:39... Y'all aren't getting it. Sora isn't the main product. Sora is proof that AGI is coming. THAT is what's going to replace all these jobs. This is literally the worst this tech will ever be. It's in it's absolute infancy and it's videos are almost indistinguishable from reality.
everyone understand that it wont be worse than the best they show, that's how progress works but the amount of compute it's required may halt the expanse due to cost.
The average shot length is 2-4 seconds in a modern film, so that's 2 minutes for a 2 second shot in Sora. That's not bad at all, especially if it makes less mistakes. You could control the consistency and the end result to an extent one imagines by using image to video in Sora. So, to that end, Sora seems very useable for producing actual full-length content.
Sora’s rendering time is slow right now because they are not using AI specific chips, but I think in a really short time they are going to make big jump with AI chips and might even take a few minutes to render 5 to 10 minutes of video from Sora, just check out Groq AI chip and how fast and efficient is it LLMS 🤯
Workflow is King in a creative process. 😀 Suno is king now and starts to overshadows a lot in the "real music production" world. The ability to create consistent comic strips (storyboards) might be an important step into making consistent movies.
Totally. We’re teaching Hollywood what the democratized future of storytelling looks like. We’re not about keeping these tools only for the industry alone.
LMAO. A year ago the technology was garbage. Now it's indistinguishable from reality, but slow. A year MAYBE two from now it'll be indistinguishable from reality and rendering almost in real time. This is the absolute worst this tech is ever gonna be and it makes me laugh when I see people use the logic of "it's imperfect now, therefore it's gonna be a long long time before it gets better". You same people were calling Chat GPT overhyped last year. It's all but ended the writing and editing industry (they can postpone it for the moment with their strikes, but the tech will win in the end) and Sora will be the death of Hollywood, Netflix, Disney, you name it.
Ya'll trippin. Look at how much AI generated videos improved in just less than a year. 2-3 years from now you will have ai generated movies. Unfortunately. I hate it.
This is by far the best YT channel to keep up to date and informed about everything AI! Updates, honest reviews, new tools, everything! Thank you and your team for all of your hard work 🙏🏻 Keep up the amazing videos! I have recommended this channel to all of my students.
about the sora generation times: 1) if it takes an hour for a 1 minute vid... thats only a minute rendering per second of of video... 2) who is actually gonna regularly generate minute long clips... when i make stuff every cut is like 4 or 5 seconds max 3) you can parallelize your gens and do like 5 at a time and have them all pop out 5 minutes later and look them over... you're making it out to be a much bigger deal than it really is... its barely slower than runway on a "seconds generated / minutes rendererd" basis and the quality is SOOOO much better its well worth the wait.
Its gonna go even better there is proof that the original potential of sora ai is still kept in reserve. They just baited others thats it.... Now after everyone comes with their max they gonna nail them
There is the fact also that usually films are done using pre-renders with lower quality, final rendera are only done way after. For example you could easily skip some harder stuff to do, hair, mirror effects, clothing, even sharp skin textures.
Don't know if ur going to see this - but make a video about how to have continuity and character consistency to REALLY tell a history and not be another AI random-nonsense film.
It's true, human emotion and storytelling is what makes films good. Not just slow-mo images. With that being said, please see our gallery and you'll see that we curate stories each week. I think you'll be surprised!
worth noting (as a vfx artist) that 1 hour to see a 1 minute video clip is very little in the filmmaking world. an average drama shoots 5 mins of screen-time in a day. FX work, you're looking at seconds produced per week. it's not fast enough for TH-camrs, but it's certainly fast enough for Hollywood, and Hollywood has the cash.
One hour for one minute is nothing. An average shot in vfx is seconds. You can potentially make 20/25 shots per hour. That's insane. Another aspect that is incredible is that the results don't look CGI like most of the full CGI shots in vfx. Sora is really a game changer.
True, legacy methods in comparison doesn't sound so bad. It's crazy how expectations are widely much faster for the AI products already. The future is exciting!
And look at all these other AI tools that have been released in the last few years... they all get better over time. But that's the worst it will likely ever be. Is it perfect for creating a feature film right now? No. But who knows where it will be 2 years from now or 5.
My repect, i having a go at midjourney 6 to see if i can keep my characters consistent. I seem to have it working. Next is the scenes and back grounds. A bridge of a star ship should be a good test. ty again friend
I don't know what the FPS is on Sora videos but I would be overjoyed to get a second per minute on most of my renders. Renders can easily take 30 minutes to get a clean output and that's for some pretty standard stuff. A frame can easily take hours or days depending on scene complexity and if you have things like caustics to worry about. I don't even really work with my image generations immediately, I know that 9/10 of them will be bad so I go do something else while it generates 100. Treat it like you're generating thousands of images because that's what animation is and it doesn't seem so unreasonable. As for not being able to direct the scene in more detail, I don't count that as a knock against it because we knew it wouldn't have that, though it was good to hear it will have image to video capabilities so that does allow a fair amount of control over the overall scene, if not the motion itself.
Completely different question by the way. With all the tools out there at the moment (except Sora of course), to your knowledge is there actually a project that is trying to make a full-length AI movie ? Is anyone aware of this ?
It takes an hour to create a one minute clip TODAY. That's going to improve as more compute is added. And the cost of that additional computing power will continue to fall, unlike the cost of people and real-world production.
The cost of computing is likely to go up, it’s getting harder to make cost, we are at 3nm chip, quantum tunneling is at 1nm. At the most we will get an 2x to 3x before we hit the physics limit.
The nudge is that the ref armors the cull, and the craft is a board, a bud in the bevy, and a rincon for the phonic. Your flow, the interest, and the intrude for the goblet, the gauge, and the grill you put are to be the veil, the vest, and the very of the solvent, the silt, and the sounding.
Thanks for featuring our feature film premiere happening on March 6th. See you there! 🎉😎 2:38
💀💀💀💀💀
🎉
🔥
😂 I’ll be there
Can't wait! It's going to be AWESOME!
1:39... Y'all aren't getting it. Sora isn't the main product. Sora is proof that AGI is coming. THAT is what's going to replace all these jobs.
This is literally the worst this tech will ever be. It's in it's absolute infancy and it's videos are almost indistinguishable from reality.
everyone understand that it wont be worse than the best they show, that's how progress works but the amount of compute it's required may halt the expanse due to cost.
Interesting perspective!
ahhaah that beginning :D u guys are awesome :D
LOL we're happy you enjoyed that bit !
The average shot length is 2-4 seconds in a modern film, so that's 2 minutes for a 2 second shot in Sora. That's not bad at all, especially if it makes less mistakes. You could control the consistency and the end result to an extent one imagines by using image to video in Sora. So, to that end, Sora seems very useable for producing actual full-length content.
Fair point!
1 - Stable Dff.
2 - Dall E
3 - Midjourney
4 - Leonardo
Not first, but that's OK since I got to open the show!
Haha you still found a way to be first.
1.) Stable Diffusion 2.) Leonardo 3.) MidJourney 4.) DALL-E
Funtimes! When is the next filmmakers event you mentioned? Mind dropping the ticket link?
Check out our posted events at our website :) www.curiousrefuge.com
@@curiousrefugeWill do thanks!
1. Stable Diffusion
2. Dall-E
3. Leonardo
4. Midjourney
1. Stable Diffusion 3
2. Dall-E 3
3. Midjourney
4. Leonardo
Image no 1 - stable diffusion
Image no 2 - leonardo ai
Image no 3 - dall-e
Image no 4 - midjourney
Sora’s rendering time is slow right now because they are not using AI specific chips, but I think in a really short time they are going to make big jump with AI chips and might even take a few minutes to render 5 to 10 minutes of video from Sora, just check out Groq AI chip and how fast and efficient is it LLMS 🤯
Well 1h for 1mn, that makes 4mn for 4s ! not that far from actual rendering time for usual 4sec videos from Runway…
@@genmakers yes good point 👍 I’m not to familiar with runway but I believe they don’t render to 1080p right?
5-10 minutes for 1 minute clip would be interesting!
1 - Stable Diffusion
2 - Dall E
3 - Leonardo
4 - Midjourney
Image 1. Stable Diffusion 3
Image 2. Midjourney
Image 3. Leonardo
Image 4. Dall-E
1. Dall e
2. SD
3. Leonardo
4. MJ
haha TIM WHAT ARE YOU DOING?!
Tim is doing Tim! He's be the best :)
Workflow is King in a creative process. 😀 Suno is king now and starts to overshadows a lot in the "real music production" world. The ability to create consistent comic strips (storyboards) might be an important step into making consistent movies.
Suno is definitely one to watch out for in the next year or two. The music industry is definitely about to change!
I thought AI has been writing Disney's scripts for years? that sh*t cant be handcrafted? [michael scott cry] no no NOOOOOOOOOO!!!
LOL sometimes it may seem like that, huh! However I'm not sure that is the case yet :)
As a musician/ artist, to see this all these new AI tools is both amazing and depressing.
It will certainly be disruptive!
Image No.1 is Stable Diffussion
Image No. 2 is Dall-e
Image No. 3 is Midjourney
Image No.4 is Leonardo
Shouldn't the democratization of content start with everyone? Not just the most "imprortant successful people in hollywood"
It is just for testing. In a couple of months it will be available for all of us.
He's teaching classes to everyone too. Reaching out to Hollywood is a way to both warn them and maybe keep artists who work for them informed.
Totally. We’re teaching Hollywood what the democratized future of storytelling looks like. We’re not about keeping these tools only for the industry alone.
@@MartinZanichelli testing is only opened to few.
Your mom does not use that software, sir.
Indeed she does!
I KNEW the Sora hype wasn’t justified! Great highlight vid
LMAO. A year ago the technology was garbage. Now it's indistinguishable from reality, but slow. A year MAYBE two from now it'll be indistinguishable from reality and rendering almost in real time.
This is the absolute worst this tech is ever gonna be and it makes me laugh when I see people use the logic of "it's imperfect now, therefore it's gonna be a long long time before it gets better". You same people were calling Chat GPT overhyped last year. It's all but ended the writing and editing industry (they can postpone it for the moment with their strikes, but the tech will win in the end) and Sora will be the death of Hollywood, Netflix, Disney, you name it.
Thanks for watching!
True, it's insane how fast this tech is progressing!
🎉🎉🎉🎉🎉
Thanks for watching!
Ya'll trippin. Look at how much AI generated videos improved in just less than a year. 2-3 years from now you will have ai generated movies. Unfortunately. I hate it.
Don't hate...appreciate! :)
This is by far the best YT channel to keep up to date and informed about everything AI!
Updates, honest reviews, new tools, everything!
Thank you and your team for all of your hard work 🙏🏻 Keep up the amazing videos!
I have recommended this channel to all of my students.
Hello AI, bye-bye Blender, Houdini, Unreal Engine etc given time lol! Best AI news channel on the Web, full-stop.
Thanks so much for watching. Glad you enjoyed this!
@@curiousrefuge I really did thank you.
about the sora generation times: 1) if it takes an hour for a 1 minute vid... thats only a minute rendering per second of of video... 2) who is actually gonna regularly generate minute long clips... when i make stuff every cut is like 4 or 5 seconds max 3) you can parallelize your gens and do like 5 at a time and have them all pop out 5 minutes later and look them over... you're making it out to be a much bigger deal than it really is... its barely slower than runway on a "seconds generated / minutes rendererd" basis and the quality is SOOOO much better its well worth the wait.
That article is not accurate. Sam Altman on Twitter was generating videos in less than 10 minutes for people who asked on X (Twitter)
Groq LPU. 20x faster than nividia GPU
Its gonna go even better there is proof that the original potential of sora ai is still kept in reserve. They just baited others thats it.... Now after everyone comes with their max they gonna nail them
There is the fact also that usually films are done using pre-renders with lower quality, final rendera are only done way after.
For example you could easily skip some harder stuff to do, hair, mirror effects, clothing, even sharp skin textures.
“Who is gonna generate minute long clips”. Me! Film editors who want options to pick the best part. Or ramp things up etc.
The issue with Gemini image generation was opposite to what is suggested here - it wasn't historically accurate, or accurate to reality.
The most needless pandering since love action beauty and the beast
Sorry!
True, Gemini has a lot of fixing to do to rebuild trust.
1 - SD3
2- Leonardo
3- Midjourney
4- Dall-E
My guess also, This is the way
1 - Stable Dff.
2 - Dall E
3 - Leonardo
4 - Midjourney
1. Stable D
2. DALL-E
3. Midjourney
4. Leonardo
Kind of..?
I would say: 1 SD3 // 2 DALL E // 3 Leonardo // 4 Midjourney 👩🏻🎓
1 - Stable Diffusion 3
2 - Dall E
3 - Midjourney
4 - Leonardo
#1 SD3, #2 dalle3, #3 midjourney, #4 Leonardo
My brain is allergic to Discord. MJ website needs to be accessible. ❤
Don't know if ur going to see this - but make a video about how to have continuity and character consistency to REALLY tell a history and not be another AI random-nonsense film.
It's true, human emotion and storytelling is what makes films good. Not just slow-mo images. With that being said, please see our gallery and you'll see that we curate stories each week. I think you'll be surprised!
New V3 mode in SUNO got copyright restrictions. You can’t use any lyrics from any song.
1. SD3
2. DALL E
3. Leonardo
4. Midjourney
worth noting (as a vfx artist) that 1 hour to see a 1 minute video clip is very little in the filmmaking world. an average drama shoots 5 mins of screen-time in a day. FX work, you're looking at seconds produced per week. it's not fast enough for TH-camrs, but it's certainly fast enough for Hollywood, and Hollywood has the cash.
Really interesting point, Christian!
1 - Stable Diffusion.
2 - Dall E
3 - Leonardo
4 - Midjourney
1. Stable Diffusion 3
2. Dall E
3. Leonardo
4. Midjourney
1. Leonardo
2. DALL-E
3. Midjourney
4. Stable Diffusion
One hour for one minute is nothing. An average shot in vfx is seconds. You can potentially make 20/25 shots per hour. That's insane. Another aspect that is incredible is that the results don't look CGI like most of the full CGI shots in vfx. Sora is really a game changer.
True, legacy methods in comparison doesn't sound so bad. It's crazy how expectations are widely much faster for the AI products already. The future is exciting!
And look at all these other AI tools that have been released in the last few years... they all get better over time. But that's the worst it will likely ever be. Is it perfect for creating a feature film right now? No. But who knows where it will be 2 years from now or 5.
The tools are advancing so fast it's difficult to know what 5 years will look like!
1. DALL-E
2. SD3
3. Midjourney
4. Leonardo
Really wish Apple had partnered with Eleven Labs. Tim Cook seems to be asleep at the wheel on AI.
Lots of people say Tim is quietly staying ahead but refusing to use the words "AI". We'll see!
My repect, i having a go at midjourney 6 to see if i can keep my characters consistent. I seem to have it working. Next is the scenes and back grounds. A bridge of a star ship should be a good test. ty again friend
1: Stable D
2: Leo
3: Mid-J
4: Dall-E
Giving this a shot... Image 1 is Stable Diffusion, Image 2 is DallE, Image 3 is Midjourney and Image 4 is Leonardo
Mum's the word.
Bird is the word!
1) stable diffusion 3
2) Dall-E 3
3) Midjourney
4) Leonardo
Hola amigos! 1 Stable Diffusion, 2 DallE, 3 Midjourney, 4 Leonardo
CRAZYYYY WHOAH
Right?!
The first image you showed of the sphere/cube/triangle/cat/dog is the image Emad tweeted out that was done with SD3 -
Thanks for the note!
I don't know what the FPS is on Sora videos but I would be overjoyed to get a second per minute on most of my renders. Renders can easily take 30 minutes to get a clean output and that's for some pretty standard stuff. A frame can easily take hours or days depending on scene complexity and if you have things like caustics to worry about. I don't even really work with my image generations immediately, I know that 9/10 of them will be bad so I go do something else while it generates 100. Treat it like you're generating thousands of images because that's what animation is and it doesn't seem so unreasonable.
As for not being able to direct the scene in more detail, I don't count that as a knock against it because we knew it wouldn't have that, though it was good to hear it will have image to video capabilities so that does allow a fair amount of control over the overall scene, if not the motion itself.
Appreciate you watching and adding your thoughts!
Completely different question by the way. With all the tools out there at the moment (except Sora of course), to your knowledge is there actually a project that is trying to make a full-length AI movie ? Is anyone aware of this ?
There is a Terminator 2 parody movie coming out that has a runtime of about 80minutes!
1 - Stable Diffuson
2- Midjourney
3- Leonardo
4- Dall-E
1 - Stable Diffusion
2 - Dall E
3 - Midjourney
4 - Leonardo
1.) Stable Diffusion 3
2.) Dall-E 3
3.) Midjourney
4.) Leonardo
1-SD
2-DALL-E3
3- Midjourney
4-Leonardo
It was so great to have you. Thank you for all your insights!
Thank YOU! :)
It takes an hour to create a one minute clip TODAY. That's going to improve as more compute is added. And the cost of that additional computing power will continue to fall, unlike the cost of people and real-world production.
The cost of computing is likely to go up, it’s getting harder to make cost, we are at 3nm chip, quantum tunneling is at 1nm. At the most we will get an 2x to 3x before we hit the physics limit.
It's true about the time but we're very excited to see the enhancements and progress over the next year.
Suno is straight up blowing my mind right now.
Right?!
Thanks for the OurT2Remake Shout out! Hope to see you there.
You bet! Caleb/Shelby will be there :)
🤘@@curiousrefuge
great backing track 🤣
What do you mean?!
@@curiousrefuge you said there was backing music but I couldn’t hear anything? Maybe it’s just my speakers 🤷🏼♂️
Leonardo
DallE
Stable Diffusion
Midjourney
fourth !
Nice job!
The nudge is that the ref armors the cull, and the craft is a board, a bud in the bevy, and a rincon for the phonic. Your flow, the interest, and the intrude for the goblet, the gauge, and the grill you put are to be the veil, the vest, and the very of the solvent, the silt, and the sounding.
What????
@@curiousrefuge 9:29 😁