Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Veo 3 and Imagen 4, and a new tool for filmmaking called Flow

blog.google

Post reply on HN
Print view View original

561–570 of 570 posts

Re: Veo 3 and Imagen 4, and a new tool for filmmaking called Flow

#561

Post by quantumHazer »

onlyrealcuzzo wrote: post #65
quantumHazer wrote: post #8

Like most AI image or video generation tools, they produce results that look good at first glance, but the more you watch, the more flaws and sloppiness you notice, and they really lack storytelling

AI used to be quite bad at coding just - what - 2 years ago? Now it's "good enough" for a lot of cases (and the pace of improvement is astounding). AI is still not great at image gen and video gen, but the pace of improvement is impressive. I'm skeptical image, video, and sound gen are "too difficult" for AI to get "good enough" at for many use cases within the next 5 years.

it's still quite bad though

Re: Veo 3 and Imagen 4, and a new tool for filmmaking called Flow

#562

Post by quantumHazer »

Closi wrote: post #504
horhay wrote: post #435

Earlier quoted context omitted.

In the era of more software advancements, people have perceived the progress of tech to be incremental more than the opposite. I think even tech bigwigs of the past have said extrapolation is a futile endeavor because technology is unpredictable.

Bill Gates saw computers and extrapolated that there would be one in every home, Gordon Moore produced Moores Law which literally extrapolated compute in the 60s and almost still holding, and you can extrapolate that GPT-5 will perform better than GPT-4, and that there will probably be a GPT-6 that will be even better than that. Extrapolating that technology will get better in the future when it has got better in the…

No, past performances are not an indicator of future performance. Sometimes it happens to be, but many other times it does not.

Re: Veo 3 and Imagen 4, and a new tool for filmmaking called Flow

#563

Post by Closi »

quantumHazer wrote: post #562
Closi wrote: post #504

Earlier quoted context omitted.

Bill Gates saw computers and extrapolated that there would be one in every home, Gordon Moore produced Moores Law which literally extrapolated compute in the 60s and almost still holding, and you can extrapolate that GPT-5 will perform better than GPT-4, and that there will probably be a GPT-6 that will be even better than that. Extrapolating that technology will get better in the future when it has got better in the…

No, past performances are not an indicator of future performance. Sometimes it happens to be, but many other times it does not.

I think this is a strange take - Technological advancements aren't mutual funds.

Of course it's an indicator of future performance - Not a guarantee, but certainly a indictator.

Re: Veo 3 and Imagen 4, and a new tool for filmmaking called Flow

#564

Post by dsadfjasdf »

jplusequalt wrote: post #557
I-M-S wrote: post #556

Earlier quoted context omitted.

Wouldn't all this hold true for AI videos as well, it's just that you'll be going through hundreds of prompts and bad generations before you get to something approaching your vision? N.B. If directing people is the distinguishing factor, then animation directors are robbed of their claim to artistry as well, as is basically any solo artist in history

I never said directing people is the distinguishing factor. I only said a director is an artist as they are deeply involved in the creation of the art in question. There is no hidden layer of linear algebra that obfuscates the creative process and that does the majority of the work.

this view is really limited. like I said, people who will do the work... will. You are thinking of the majority of lazy people who aren't producing anything of real value in the design/art world already.

Re: Veo 3 and Imagen 4, and a new tool for filmmaking called Flow

#565

Post by horhay »

Closi wrote: post #504
horhay wrote: post #435

Earlier quoted context omitted.

In the era of more software advancements, people have perceived the progress of tech to be incremental more than the opposite. I think even tech bigwigs of the past have said extrapolation is a futile endeavor because technology is unpredictable.

Bill Gates saw computers and extrapolated that there would be one in every home, Gordon Moore produced Moores Law which literally extrapolated compute in the 60s and almost still holding, and you can extrapolate that GPT-5 will perform better than GPT-4, and that there will probably be a GPT-6 that will be even better than that. Extrapolating that technology will get better in the future when it has got better in the…

Yeah, no. That's not my point. Saying something will get better in /any/ amount isn't unreasonable. That's being positive towards innovation and it's harmless. That's not exactly what the modern tech consumer is being made to expect nowadays. It's always the next iPhone moment. The next GPT. Always a huge leap. Like if we landed on the moon and we've been promised we'll have colonies outside the solar system by 2040

Re: Veo 3 and Imagen 4, and a new tool for filmmaking called Flow

#566

Post by horhay »

Closi wrote: post #563
quantumHazer wrote: post #562

Earlier quoted context omitted.

No, past performances are not an indicator of future performance. Sometimes it happens to be, but many other times it does not.

I think this is a strange take - Technological advancements aren't mutual funds. Of course it's an indicator of future performance - Not a guarantee, but certainly a indictator .

I think what I genuinely have an issue with is that extrapolation ignores the finer details of research work being taken to understand the current issues that may affect the future goals of a technology. It basically just ignores the science of it in favor of "well we've seen this level work before". The thing is, sometimes the more these experts learn about the current goal to beat in their product, the more things previously unknown and unplanned for present themselves as new issues. Extrapolating is a bit past-facing when you see it this way, because those new problems are entirely new challenges that haven't been considered in the equation and may require an entirely new approach. Every success is different, every failure, and every challenge is different. So extrapolating seems pointless.

Re: Veo 3 and Imagen 4, and a new tool for filmmaking called Flow

#567

Post by teitoklien »

sensanaty wrote: post #379
teitoklien wrote: post #235

Earlier quoted context omitted.

> but the consequences are the opposite of supporting the diversity and abundance in arts. I made 5 songs about 5 different people in a week, with carefully crafted lyrics and tones described by me in the custom prompt, that led to 10 mp3 files of songs (suno generates 2 songs per prompt) Those songs are out there, it’s different, it’s not sloppy it’s actually quite enjoyable. Now there is more diversity and abundanc…

I'm sorry but that song you posted is the most generic, bland garbage that imitates the literally millions of other pop songs they stole in their training sets. It sounds like Charlie xcx without any of the actual creativity, and the actual track itself you could learn how to make yourself with FL Studio in about half an hour following one of the many tutorials out there.

if a song style gets imitated a million times that speaks to how good a particular track is, people dont imitate trash.

Also it took me 30 seconds to generate it, instead of 7 days of learning and debugging sound studio tools (which i dont have time for, i need to work daily too)

So No, you’re only speaking from emotions, not from a place of reality.

Re: Veo 3 and Imagen 4, and a new tool for filmmaking called Flow

#568

Post by jbkkd »

anilgulecha wrote: post #497
_xhhf wrote: post #418

Earlier quoted context omitted.

It's not a theory, at Cannes a feature movie has premiered that is generated entirely by AI. Made in Spain.

can you share a link/details, please?

https://ground.news/article/the-great-reset-the-first-photor...

Re: Veo 3 and Imagen 4, and a new tool for filmmaking called Flow

#569

Post by vunderba »

andybak wrote: post #397
andybak wrote: post #395

Earlier quoted context omitted.

Side note. It's my understanding that being a pith helmet is pretty orthogonal to having a spike. Plenty of helmets with spikes aren't pith helmets and plenty of pith helmets don't have spikes. Not sure if this affects your results or not but I resist chiming in!

Also "Hippity Hop" is a Space Hopper! Wikipedia agrees with me: https://en.wikipedia.org/wiki/Space_hopper :) I wonder how much the commonality or frequency of names for things affects image generation? My hunch is that it it roughly correlates and you'd get better results for terms with more hits in the training data. I'd probably use Google image search as a rough proxy for this.

well I'm about a billion years late to this but going to reply anyway. It's not mentioned but internally I give each model several "iterations" of the prompts themselves. This included using hippity hop, space hopper, and even a physical description of the toy itself. But it's a good call out!

Re: Veo 3 and Imagen 4, and a new tool for filmmaking called Flow

#570

Post by tintor »

viraptor wrote: post #268
tintor wrote: post #140

Earlier quoted context omitted.

More difficult examples: - wine glass that is full to the edge with wine (ie. not half full) - wrist watch not showing V (hands at 10 and 2 o'clock) - 9 step IKEA shelf assembly instruction diagram - any kind of gymnastics / sport acro

What's the reason to test the "not showing ..."? I've never seen anyone make that kind of request in real life. They ask for what they actually want instead. You'd ask for a clock showing 3:25 rather than "not 10:10". I mean, it's a fun edge case, but I'm practice - does it matter?

Problem is that watchmakers always set the watches to show V with clock hands when they market their watches. This causes a very strong bias in image generation models, making it very difficult for them to generate watch that shows any other time, even if user requests it.
Post reply on HN