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

Veo 2: Our video generation model

deepmind.google

Post reply on HN
Print view View original

141–150 of 342 posts

Re: Veo 2: Our video generation model

#141

Post by achierius »

EGreg wrote: post #43
turnsout wrote: post #14

Earlier quoted context omitted.

Put another way, over time people devalue things which can be produced with minimal human effort. I suspect it's less about humanity's values, and more about the way money closely tracks "time" (specifically the duration of human effort).

I strongly disagree. How many clothes do you buy that have 100 thread count, and are machine-made, vs hand-knit sweaters or something? When did you ask people for directions, or other major questions, instead of Google? You can wax poetic about wanting "the human touch", but at the end of the day, the market speaks -- people will just prefer everything automated. Including their partners, after your boyfriend can rem…

>PS: anything you write on HN can already have been written by AI

Yeah in some broad sense, the same as we've always had: back in the 2010s it could have been generated by a Markov chain, after all. The only difference now is that the average quality of these LLMs is much, much higher. But the distribution of their responses is still not on par with what I'd consider a good response, and so I hunt out real people to listen to. This is especially important because LLMs are still not capable of doing what I care most about: giving me novel data and insights about the real world, coming from the day to day lived experience of people like me.

HN might die but real people will still write blogs, and real people will seek them out for so long as humans are still economically relevant.

Re: Veo 2: Our video generation model

#142

Post by jerpint »

simonw wrote: post #131

I got access to the preview, here's what it gave me for "A pelican riding a bicycle along a coastal path overlooking a harbor" - this video has all four versions shown: https://static.simonwillison.net/static/2024/pelicans-on-bic... Of the four two were a pelican riding a bicycle. One was a pelican just running along the road, one was a pelican perched on a stationary bicycle, and one had the pelican wearing a weird…

It looks much better than Sora but still kind of in uncanny valley

Re: Veo 2: Our video generation model

#143

Post by yieldcrv »

gamesbrainiac wrote: post #56

This might be a dumb question to ask, but what exactly is this useful for? B-Roll for YouTube videos? I'm not sure why so much effort is being put into something like this when the applications are so limited.

this is perfect for the landing page of any website I make

my templates all are waiting for stock videos to be added looping in the background

you have no idea how cool I am with the lack of copyright protections afforded to these videos I will generate, I'm making my money other ways

Re: Veo 2: Our video generation model

#144

Post by griomnib »

echelon wrote: post #139
dyauspitr wrote: post #105

Earlier quoted context omitted.

The quality for SD is no where near the clear leaders.

> The quality for SD is no where near the clear leaders. It absolutely is. Moreover, the tools built on top of SD (and now Flux) are superior to any commercial vertical. The second-place companies and research labs will continue to release their models as open source, which will cause further atrophy to the value of building a foundation model. Value will accrue in the product, as has always been the case.

SD will also generate what I tell it, unlike the corporate models that have all kinds of “safeguards”.

Re: Veo 2: Our video generation model

#145

Post by seabombs »

I'm always curious with the examples in these announcements, how close is the training data to the sample prompts? And how much of the prompt is important or ends up ignored in the result?

The prompt for the figure running through glowing threads seems to contain a lot of detail that doesn't show up in the video.

In the first example (close-up of DJ), the last line about her captivating presence and the power of music I guess should give the video a "vibe" (compared to prescriptively describing the video). I wonder how the result changes if you leave it out?

Cynically I think that it's a leading statement there for the reader rather than the model. Like now that you mention it, her presence _is_ captivating! Wow!

Re: Veo 2: Our video generation model

#146

Post by nope96 »

gamesbrainiac wrote: post #56

This might be a dumb question to ask, but what exactly is this useful for? B-Roll for YouTube videos? I'm not sure why so much effort is being put into something like this when the applications are so limited.

This is a first step towards "the holodeck". You describe a scene and it exists. Imagine you could jump in and interact with it. That seems like something that could happen in 10-20 years.

Re: Veo 2: Our video generation model

#150

Post by throw4321 »

visnup wrote: post #109
jasonjmcghee wrote: post #45

I appreciate they posted the skateboarding video. Wildly unrealistic whenever he performs a trick - just morphing body parts. Some of the videos look incredibly believable though.

our only hope for verifying truth in the future is that state officials give their speeches while doing kick flips and frontside 360s.

What officials actually say doesn't make a difference anymore. People do not get bamboozled because of lack of facts. People who get bamboozled are past facts.
Post reply on HN