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

Stable Video Diffusion

stability.ai

Post reply on HN
Print view View original

311–316 of 316 posts

Re: Stable Video Diffusion

#311

Post by whywhywhywhy »

pegasus wrote: post #294
whywhywhywhy wrote: post #286

Earlier quoted context omitted.

Nah I disagree, this feels like a glorification of the process not the end result. Just because having the 3D model in the scene with all the lighting makes the end result feel more solid to you because you feel you can see the work that's going into it. In the end diffusion technology can make a more realistic image faster than a rendering engine can. I feel pretty strongly that this pipeline will be the foundation…

Nah, I agree with GP. Who didn't suggest making 3D scenes by hand, but the opposite: create those 3D scenes using the generative method, use ray-tracing or the like to render the image. Maybe have another pass through a model to apply any touch-ups to make it more gritty and less artificial. This way things can stay consistent and sane, avoiding all those flaws which are so easy to spot today.

I know exactly what OP suggested but why are you both glorifying the fact there is a 3D scene graph made in the middle and then slower rendering at the end when the tech can just go from the first thing to a better finished thing?

Re: Stable Video Diffusion

#312

Post by ronsor »

yorwba wrote: post #259
frognumber wrote: post #226

Earlier quoted context omitted.

There are a few problems: 1) You and I invent our own private "copyright" for data (which is not copyrightable) 2) Everything is fine until my wife walks up to my computer and makes a copy of the data. She's not bound by our private "copyright." She doesn't even know it exists, and shares the data with her bestie. And... our private pseudo-copyright is dead. Also: Licenses are not the same as contracts. There are tim…

> my wife walks up to my computer and makes a copy of the data As you agreed to in our contract, you now need to compensate me for the damage caused by your failure to prevent unauthorized third-party access. Of course you're free to attempt to recover the sum you have to pay me from your wife. > The output of a program is rarely copyrightable by the author (as opposed to the user). The author of the program can make…

> The author of the program can make it a condition of letting the user use the program that the user has to assign all copyright to the author of the program

I think it is quite likely a court would find that unconscionable.

Re: Stable Video Diffusion

#313

Post by pegasus »

whywhywhywhy wrote: post #311
pegasus wrote: post #294

Earlier quoted context omitted.

Nah, I agree with GP. Who didn't suggest making 3D scenes by hand, but the opposite: create those 3D scenes using the generative method, use ray-tracing or the like to render the image. Maybe have another pass through a model to apply any touch-ups to make it more gritty and less artificial. This way things can stay consistent and sane, avoiding all those flaws which are so easy to spot today.

I know exactly what OP suggested but why are you both glorifying the fact there is a 3D scene graph made in the middle and then slower rendering at the end when the tech can just go from the first thing to a better finished thing?

Because it just can't. And it won't. It can't even reliably produce consistent shadows in a still image, so when we talk video with a moving camera, all bets are off. To create flawless movie simulations through a dynamic and rich 3D world, requires an ability of internally represent that scene with a level of accuracy which is beyond what we can hope generative models to achieve, even with the gargantuan amount of GPU-power behind ChatGPT, for example. ChatGPT, may I remind you, can't even properly simulate large-ish multiplications. I think you may need to slightly recalibrate your expectations for generative tech here.

Re: Stable Video Diffusion

#314

Post by zvictor »

rbhuta wrote: post #224

VRAM requirements are big for this launch. We're hosting this for free at https://app.decoherence.co/stablevideo . Disclaimer: Google log-in required to help us reduce spam.

it's worth paying your subscription just for these free videos. would those have the watermark removed if I go "Basic"?

Re: Stable Video Diffusion

#315

Post by bbor »

dragonwriter wrote: post #300
bbor wrote: post #298

Earlier quoted context omitted.

I find that very unlikely. LLMs seem capable of simulating human intuition, but not great at simulating real complex physics. Human intuition of how a scene “should” look isn’t always the effect you want to create, and is rarely accurate im guessing

> LLMs seem capable of simulating human intuition, but not great at simulating real complex physics. Diffusion models aren't LLMs (they may use something similar as their text encoder layer) and they simulate their training corpus, which usually isn't selected solely for physical fidelity, because that's not actually the single criteria for visual imagery outside of what is created by diffusion models.

Huh fair enough. I mean they are large models based on language but I see your point. Even though everything you said is true, I still believe there’s a place for human-constructed logically-explicit simulations and functions. In general, and in visual arts.

Re: Stable Video Diffusion

#316

Post by achileas »

btbuildem wrote: post #287
achileas wrote: post #216

Earlier quoted context omitted.

> Has anyone come across a solution where model can iterate (eg, with prompts like "move the bicycle to the left side of the photo")? It feels like we're close. Nearly all of the available models have this, even the highly commercialized ones like in Adobe Firefly and Canva, it’s called inpainting in most tools.

I think that's more "inpainting" where the existing software solution uses AI to accelerate certain image editing tasks. I was looking for whole-image manipulation at the "conceptual" level.

They have this. Inpainting is just a subset of the image-to-image workflow and you don't have to provide a region if you want to do whole-image manipulation.
Post reply on HN