Google VEO2 AI-created Porsche SPEC AD.

[Ty Harper]

Stunning.

[MrSMW]

Nice, but I don't get the AI bit?

I was expecting an entire ad created with just AI and was watching speechless...but the BTS revealed a 110 crew, what looked like a proper actors/filming/crew job and production to me?

Unless the BTS was also AI? 😉

[BTM_Pix]

1 hour ago, MrSMW said:

Unless the BTS was also AI? 😉

Indeed it was.

 

[MrSMW]

33 minutes ago, BTM_Pix said:

Indeed it was.

 

Really?

The ad and the BTS were both fully AI?

Nothing real about any of it?

If so, I am so out of touch with this world and the entire industry is fucked, - get out now 😵

[BTM_Pix]

17 minutes ago, MrSMW said:

Really?

The ad and the BTS were both fully AI?

The voiceover was real.

This is the card at the end of the video.

 

18 minutes ago, MrSMW said:

Nothing real about any of it?

Well, there’s the rub.

Whatever it was trained on was certainly very real.

Looking at this frame, I’m pretty sure if you went through enough of PotatoJet’s YouTube content you’d find the ,shall we say, “inspiration” for it.

[MrSMW]

I wonder just how close we are to movie level now then...

But I suppose the material still needs to exist to pull together to make this kind of thing and as a commercial and BTS, enough exists, but not for any kind of real movie that would require a unique script and acting.

At least not without going CGI.

It's crazy how good this stuff is becoming and if I was a youngster, I certainly would not be looking at the more traditional aspects of this industry such as cameras, lenses, lighting, sound etc, but what skills I could develop in my parents basement from my laptop.

But for me personally, that would be a hard no as it would not interest me and I'd rather be a lumberjack or something.

[eatstoomuchjam]

It's definitely improved for very short takes.  While I didn't watch the whole thing, the bits I skipped around to see were all very short clips.  Anyplace with letters was also a give-away (the wording on the truck).  I also wonder how many "takes" they had to do of each clip to get one that looked good and where the physics were mostly realistic.  I saw a few clips that seemed in the "weird physics" space - that continues to be a major weakness of AI-generated video.

But the main thing that could still slow down the enshittification of everything through AI is that AI-generated material is still not able to be copyrighted - at least not in the US.  How long that lasts before the shitheads in Congress/Senate/Presidency "fix" it?  Unknown.

[BTM_Pix]

1 hour ago, eatstoomuchjam said:

Anyplace with letters was also a give-away

The corruption on the license plate coincidentally features a fragment of the plate of the car that images of may well have been used to train it with.

The car is a Porsche factory one that was provided to reviewers so enough images of it from different sources to train it.

 

Could just be a massive coincidence of course .....

I think a new parlour game in the upcoming years when generating this stuff becomes more readily/instantly available will be image makers entering a specific prompt describing one of their own images that they have put online and determining how much has been lifted from their work based on the output that comes back.

 

[Ty Harper]

2 hours ago, MrSMW said:

But I suppose the material still needs to exist to pull together to make this kind of thing and as a commercial and BTS, enough exists, but not for any kind of real movie that would require a unique script and acting.

Yes I'd agree that that is where we are at this moment - but obvi that will eventually change. At the very least I'd think that you won't need as much location shooting, b-roll/alternate shots, human extras, etc.

[Ty Harper]

55 minutes ago, BTM_Pix said:

I think a new parlour game in the upcoming years when generating this stuff becomes more readily/instantly available will be image makers entering a specific prompt describing one of their own images that they have put online and determining how much has been lifted from their work based on the output that comes back

Absolutely! I first saw the video via an IG acct so I used Google Images to search for the original video and all I got was what seemed like all the images the image I clipped was trained on.

[Marcio Kabke Pinheiro]

The AI was trained about overrigging a camera with Caleb Pike.

[Andrew Reid]

Potato AI.

[EduPortas]

11 hours ago, Andrew Reid said:

Potato AI.

The advertisement industry created its own problem by allowing minus 1 second cuts to be the norm in most ads.

For short stuff like most ads AI is fine.

For a feature film with long takes of real human emotions not so much.

The recent passing of Lynch should help us keep that in consideration. 

[KnightsFan]

49 minutes ago, EduPortas said:

For a feature film with long takes of real human emotions not so much.

*yet

It's hard for me to believe that 20 years from now, AI will be unable to produce videos of any type that are indistinguishable from videos shot with cameras.

On 1/23/2025 at 2:31 AM, BTM_Pix said:

Whatever it was trained on was certainly very real.

Looking at this frame, I’m pretty sure if you went through enough of PotatoJet’s YouTube content you’d find the ,shall we say, “inspiration” for it.

It's an interesting discussion how much diverse training needs to be given to a model before its output is presumed original. I mean, we're all trained on images from other artists, and at the moment the primary difference is that our brains are way less adept at permanently storing precise information. And at this point, these models have ingested more videos than any of us have seen, so the inspiration is even more spread out than a human's.

It's probably not that far off before some of these big companies start collecting training data straight from the real world, that combines binocular vision, audio, and other sensors into a combined stream, like our own.

[Ty Harper]

14 hours ago, KnightsFan said:

It's hard for me to believe that 20 years from now, AI will be unable to produce videos of any type that are indistinguishable from videos shot with cameras.

Agreed. 10 years from now even.

[EduPortas]

On 1/24/2025 at 6:59 PM, KnightsFan said:

*yet

There's always this uncanny-valley when I watch even the most realistic CGI.

After a while it becomes clear that's not real human I'm watching on the screen. Long takes make it easier to identify.

Maybe it's the natural movement of the human body we're accustomed to from millions of years.

 

[KnightsFan]

5 hours ago, EduPortas said:

There's always this uncanny-valley when I watch even the most realistic CGI.

After a while it becomes clear that's not real human I'm watching on the screen. Long takes make it easier to identify.

Maybe it's the natural movement of the human body we're accustomed to from millions of years.

 

So are you saying that AI generated video will never become indistinguishable from real video? Its still obvious in most cases right now, my point as that it will become better.

CGI is difficult to keep improving because it's largely a manual process. Some CG artists are very good, but many are not. It took decades for us to go from moving a cube across a screen to the very best CGI today, which takes enormous effort from large teams of extremely experienced and skilled people.

AI generated video has gone from nothing to this ad in what, two years? And how many people were involved? And the same AI model will continue improving, and never retire. Even without improvements to the methods of training or reinforcement learning, imagine the output when it has 10x or 100x the training data.

[Ty Harper]

19 minutes ago, KnightsFan said:

CGI is difficult to keep improving because it's largely a manual process. Some CG artists are very good, but many are not. It took decades for us to go from moving a cube across a screen to the very best CGI today, which takes enormous effort from large teams of extremely experienced and skilled people.

This has been the part I think most people continue to struggle with. We need to get used to the idea of AI achieving something, as an inevitability. Period. 

[Evgeniy85]

AI will improve forever. The economy will grow forever. Earth has infinite resources. Don't let anyone tell you otherwise!

[BTM_Pix]

The only way it won't achieve being indistinguishable from real video in the next few years will be if its actually outright banned.

Now that is something that I thought was a complete long shot until this DeepSeek thing reared its head.

Now that the usual suspects may not make their expected humungous financial and power gains from AI they might well get the cult leader that they own to try and do exactly that.

Get him to threaten it with yuuuge tariffs and/or the invasion of Knotty Ash to plunder its remaining deep mine jam butty resources etc. 

In terms of what AI means to creators of what will likely be called "legacy content" before too long, this video is a sobering look.

Basically, a man does a live stream of completely improvised music using modular synths (the very nature of which mean even he would be hard pushed to completely repeat it) and two years later it gets five copyright strikes from different artists and labels, a lot of which are for allegedly infringing copyright on pieces that they made after he performed and published the stream.

In particular, the final one - at the very least - looks to be completely AI generated from the music to the artist to the social media content.

Music is a couple of years ahead of video in being "blessed" with AI tools so expect this sort of stuff to be happening in the video world soon where someone makes a video with a prompt that uses elements of your content and similarly flawed automated copyright infringement tools flag your original video from years earlier and demand you give them an 80% revenue share for the privilege of not having your channel getting a strike.