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

Veo

deepmind.google

Post reply on HN
Print view View original

451–460 of 539 posts

Re: Veo

#451

Post by cpill »

ancientworldnow wrote: post #322
qingcharles wrote: post #282

Earlier quoted context omitted.

I can see using these video generators to create video storyboards. Especially if you can drop in a scribbled sketch and a prompt for each tile.

That sounds actively harmful. Often we want story boards to be less specific so as not to have some non artist decision maker ask why it doesn't look like the storyboard. And when we want it to match exactly in an animatic or whatever, it needs to be far more precise than this, matching real locations etc.

I guess this will give birth to a new kind of film making. Start with a rough sketch, generate 100 higher quality versions with an image generator, select one to tweak, use that as input to a video generator which generates 10 versions, coffee one to refine etc

Re: Veo

#452

Post by kmacdough »

sabellito wrote: post #431
salamo wrote: post #353

The first thing I will do when I get access to this is ask it to generate a realistic chess board. I have never gotten a decent looking chessboard with any image generator that doesn't have deformed pieces, the correct number of squares, squares properly in a checkerboard pattern, pieces placed in the correct position, board oriented properly (white on the right!) and not an otherwise illegal position. It seems to be…

Per usual the top comment on anything AI related is snark on "it can't to [random specific thing] well yet".

Tiring, but so is the relentless over-marketing. Each new demo implies new use cases and flexible performance. But the reality is they're very brittle and blunder most seemingly simple tasks. I would personally love an ongoing breakdown of the key weaknesses. I often wonder "can it X?" The answer is almost always "almost, but not a useful almost".

Re: Veo

#455

Post by danielbln »

alex_duf wrote: post #444
loudmax wrote: post #21

The videos in this demo are pretty neat. If this had been announced just four months ago we'd all be very impressed by the capabilities. The problem is that these video clips are very unimpressive compared to the Sora demonstration which came out three months ago. If this demo was announced by some scrappy startup it would be worth taking note. Coming from Google, the inventor of the Transformer and owner of the larg…

>these sample videos are underwhelming wow the speed at which we can be blasé is terrifying. 6 months ago this was not possible, and felt this was years away! They're not underwhelming to me, they're beyond anything I thought would ever be possible. are you genuinely unimpressed? or maybe trying to play it cool?

The faster the tech cycle, the faster we become accustomed to it. Look at your phone, an absolute, wondrous marvel of technology that would have been utterl and totally scifi just 25 years ago. Yet we take it for granted, as we do with all technology eventually. The time frames just compress is all, for better or for worse.

Re: Veo

#456

Post by ArchitectAnon »

I think the thing that most perturbs me about AI is that it takes jobs that involve manipulating colours, light, shade and space directly and turns them into essay writing exercises. As a dyslexic I fucking hate writing essays. 40% of architects are dyslexic. I wouldn't be surprised if that was similar or higher in other creative industries such as filmmaking and illustration. Coincidentally 40% of the prison population is also dyslexic, I wonder if that's where all the spare creatives who are terrible at describing things with words will end up in 20 years time.

Re: Veo

#457

Post by wasteduniverse »

sabellito wrote: post #431
salamo wrote: post #353

The first thing I will do when I get access to this is ask it to generate a realistic chess board. I have never gotten a decent looking chessboard with any image generator that doesn't have deformed pieces, the correct number of squares, squares properly in a checkerboard pattern, pieces placed in the correct position, board oriented properly (white on the right!) and not an otherwise illegal position. It seems to be…

Per usual the top comment on anything AI related is snark on "it can't to [random specific thing] well yet".

[dead]

Re: Veo

#458

Post by swyx »

salamo wrote: post #366
sdenton4 wrote: post #361

Earlier quoted context omitted.

This strikes me as equally "AI complete" as drawing hands, which is now essentially a solved problem... No one test is sufficient, because you can add enough training data to address it.

Yeah "AI complete" is a bit tongue-in-cheek but it is a fairly spectacular failure mode of every model I've tried.

ive been using “agi-hard” https://latent.space/p/agi-hard as a term

because completeness isnt really what we are going for

Re: Veo

#459

Post by kmacdough »

ugh123 wrote: post #219

From a filmmaking standpoint I still don't think this is impactful. For that it needs a "director" to say: "turn the horse's head 90˚ the other way, trot 20 feet, and dismount the rider" and "give me additional camera angles" of the same scene. Otherwise this is mostly b-roll content. I'm sure this is coming.

I wouldn't be so sure it's coming. NNs currently dont have the structures for long term memory and development. These are almost certainly necessary for creating longer works with real purpose and meaning. It's possible we're on the cusp with some of the work to tame RNNs, but it's taken us years to really harness the power of transformers.

Re: Veo

#460

Post by aavshr »

ArchitectAnon wrote: post #456

I think the thing that most perturbs me about AI is that it takes jobs that involve manipulating colours, light, shade and space directly and turns them into essay writing exercises. As a dyslexic I fucking hate writing essays. 40% of architects are dyslexic. I wouldn't be surprised if that was similar or higher in other creative industries such as filmmaking and illustration. Coincidentally 40% of the prison populat…

I would imagine and hope for interfaces to exist where the natural language prompt is the initial seed and then you'd still be able to manipulate visual elements through other ways.
Post reply on HN