Live data from Hacker News

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

blog.google

241–250 of 570 posts

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

#241
post #152

Earlier quoted context omitted.

Dude. Making a movie is not accessible to most people and it's EVERYONES dream. This is not even there yet, but I have a few movies I need to make and I will never get a cast together and go do it before I die. If some creatives need to take a backseat so a million more creatives can get a chance, then so be it.

Yeah, there will be so many AI generated videos that many will go unwatched. Not sure where this is heading but it's certainly an interesting future.

Nobody will watch anyone's "creations". The TV or whatever device watch on will observe what you and everyone else engage with and how you interact with it and create content for you on the fly, just like Instagram and TikTok's feeds do now.

AI creatives can enjoy the brief blip in time where they might get someone else to watch what they've created before their skills become obsolete in an exponentially faster rate just like everyone else's.

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

#242

Got a bit of an uncanny valley feeling with the owl and the old man videos. And the origami video give me a sort of sinister feeling, seemed vaguely threatening, agressive.

We've made so much progress in the last 20 years; it used to take huge teams of developers and artists and giant compute clusters and rendering time to generate uncanny valley!

Now it just takes giant compute clusters and inference time.

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

#243

After doing some testing, Imagen 4 doesn't score any higher than Imagen 3 on my comparison chart, approximately ~60% prompt adherence accuracy. https://genai-showdown.specr.net

I'm curious why you decide to declare victory after one successful attempt, but try many times for unsuccessful models. Are you trying to measure whether a model _can_ get it right, or whether it frequently _does_ get it right? I feel like success rate is a better metric here, or at least a fixed number of trials with some success rate threshold to determine model success.

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

#244

Earlier quoted context omitted.

> when pooled at scale between all users allows people to make their own song generators with GPUs. I can assure you, local generation is to become a fringe activity, same as self-hosting web services, only worse, because the quality gap (which in case of software is often negligible) will be insurmountable. > I’d love to hear how you think the big 3 corps (Universal, Sony, etc) who own all the music almost globally…

> 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…

What that service generates for you is content, it's not art. I don't understand how you can mix one for the other. The only artistic creative process happening is you making a prompt and reiterating. You're free to disagree, but I posit it's not nearly enough to qualify.

> Tell me this song is trash and slop

As you wish.

What I argue is not about "diversity of songs", but that's this "AI" revolution strives to further the process of replacing the arts with commodity substitutes. It will be sleek, shiny, comforting, friendly and, possibly, enjoyable, but it's almost worthless in itself.

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

#245
post #100

I do find myself wondering if the people working on this stuff ever give any real thought to the impact on society that this is going to have. I mean obviously the answer is "no" and this is going to get a bunch of replies saying that inventors are not to blame but the negative results of a technology like this are fairly obvious. We had a movie two years ago about a blubbering scientist who blatantly ignored that to…

How could you possibly push back on the societal benefit of a director being able buy a vacation home in Lake Tahoe?

And what about the rest of human pyramid working under the director employed in these productions?

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

#246
post #145

Earlier quoted context omitted.

The first two paragraphs of your argument could be used to discuss whether Photography (Camera is doing most of the work) or Digital Drawing (Photoshop is doing most of the work) are art. Both things which were dismissed as not art at first but are widely accepted as an art medium nowadays.

Everyone has a phone camera, and takes photos, but not everyone is a photographer, and even photographers wouldn’t proclaim all their photos “art”. AI cannot “democratize art” any more than the camera did, until the day it starts teaching artistry to its users.

> until the day it starts teaching artistry to its users.

It almost definitely can start teaching artistry to its users, and the same people who are mad in this thread will be mad that it's taking away jobs from art instructors.

The central problem is the same and it's what Marshall Brain predicted: If AI ushers in a world without scarcity of labor of all kinds, we're going to have to find a fundamentally new paradigm to allocate resources in a reasonably fair way because otherwise the world will just be like 6 billionaire tech executives, Donald Trump, and 8 billion impoverished unemployed paupers.

And no, "just stop doing AI" isn't an option, any more than "stop having nuclear weapons exist" was. Either we solve the problems, or a less scrupulous actor will be the only ones with the powerful AI, and they'll deploy it against us.

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

#248
post #238

An indie film with poor production values, even bad acting can grip you, make you laugh and make you cry. The consistency of quality is key - even if it is poor. The directing is the red thread throughout the scenes. Anything with different quality levels interrupts your flow and breaks your experience. The problem with AI video content at this stage is that the clips are very good 'in themselves', just as LLM result…

You might want to look up NeuralViz on YouTube. 180k subscribers. They've been building out an entire cinematic universe using AI video tools. And it's by far the funniest show I've watched in years. So the claim that "let you engage beyond an individual clip will not be possible for a long time" isn't true. People are already doing it. https://www.youtube.com/@NeuralViz

This is the first time I've wanted more AI video content. Thanks for sharing.

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

#249
post #185

Earlier quoted context omitted.

> The gates are wide open for those that want to put in effort to learn. Why is effort a requirement? Why should being an artist be a viable job? Would you be against technology that makes medical doctors obsolete?

> Why is effort a requirement? That's how human brains work. People have an intrinsic need to sort, build hierarchies and prioritize. Effort spent is one of viable heuristics for these processes. > Why should being an artist be a viable job? Art itself has great value, if it weren't, museums, theaters and live shows wouldn't exist. > Would you be against technology that makes medical doctors obsolete? The analogy doe…

I think the result of an art piece is "wow that looks pretty cool / makes me feel this certain way". Human properties aren't required there. Sure, some art rests on there being human properties, but then AI wouldn't be able to replace it by definition. You can argue that AI lowers the discoverability of art that falls in the latter category, but I'd say that it's a solvable problem that can be fixed with better recommendation algorithms.

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

#250

For anyone with an access, can you ask it to make a pickup truck drive through mud? I’ve tested various different AIs and they all suck with physics and tires spinning wrong way, it is just embarrassing. Demos look amazing, but when it comes to actual use - there is none that worked for me. I guess it is all to increase “investor value”

Google posted a video of their own of an off-roader going through mud.

https://www.youtube.com/watch?v=SPF4MGL7K5I

Obviously we don't know how hand picked that is so it would be interesting to see a comparison from someone with access.

Post reply on HN