Welcome to DU! The truly grassroots left-of-center political community where regular people, not algorithms, drive the discussions and set the standards. Join the community: Create a free account Support DU (and get rid of ads!): Become a Star Member Latest Breaking News Editorials & Other Articles General Discussion The DU Lounge All Forums Issue Forums Culture Forums Alliance Forums Region Forums Support Forums Help & Search

LiberalArkie

(19,314 posts)
Mon Jun 2, 2025, 09:01 AM Jun 2025

Holy crap: AI video just took a startling leap in realism. Are we doomed?

MAY 29, 2025 12:58 PM

Last week, Google introduced Veo 3, its newest video generation model that can create 8-second clips with synchronized sound effects and audio dialog—a first for the company's AI tools. The model, which generates videos at 720p resolution (based on text descriptions called "prompts" or still image inputs), represents what may be the most capable consumer video generator to date, bringing video synthesis close to a point where it is becoming very difficult to distinguish between "authentic" and AI-generated media.

Google also launched Flow, an online AI filmmaking tool that combines Veo 3 with the company's Imagen 4 image generator and Gemini language model, allowing creators to describe scenes in natural language and manage characters, locations, and visual styles in a web interface.

Both tools are now available to US subscribers of Google AI Ultra, a plan that costs $250 a month and comes with 12,500 credits. Veo 3 videos cost 150 credits per generation, allowing 83 videos on that plan before you run out. Extra credits are available for the price of 1 cent per credit in blocks of $25, $50, or $200. That comes out to about $1.50 per video generation. But is the price worth it? We ran some tests with various prompts to see what this technology is truly capable of.

How does Veo work?

Like other modern video generation models, Veo 3 is built on diffusion technology—the same approach that powers image generators like Stable Diffusion and Flux. The training process works by taking real videos and progressively adding noise to them until they become pure static, then teaching a neural network to reverse this process step by step. During generation, Veo 3 starts with random noise and a text prompt, then iteratively refines that noise into a coherent video that matches the description.

Snip

https://arstechnica.com/ai/2025/05/ai-video-just-took-a-startling-leap-in-realism-are-we-doomed/


Comment: I love tech, I have always loved tech since I was a toddler in the 1950's. But these videos just scared the crap out of me..
If there was ever a "Must see tv" moment, then the ones in this article are them.

16 replies = new reply since forum marked as read
Highlight: NoneDon't highlight anything 5 newestHighlight 5 most recent replies
Holy crap: AI video just took a startling leap in realism. Are we doomed? (Original Post) LiberalArkie Jun 2025 OP
We are in trouble. Trust and verification of what we see is questionable. marble falls Jun 2025 #1
"Mountainhead" here we come. sinkingfeeling Jun 2025 #2
Fingers Johnny2X2X Jun 2025 #3
Not any more.. LiberalArkie Jun 2025 #6
Already addressed. IOW, there's no way for you to tell anymore 0rganism Jun 2025 #13
At least the porn will be awesome. Midnight Writer Jun 2025 #4
I say bigmonk Jun 2025 #5
Okay, who's my stalker? hunter Jun 2025 #7
It can't do text very well k_buddy762 Jun 2025 #8
Roger Stone is already limbering up his tentacles for future political ads. Buns_of_Fire Jun 2025 #9
It's impressive technology TheProle Jun 2025 #10
Wow, amazing! Thanks for posting this. Abolishinist Jun 2025 #11
I really do not see any reason why not as most of the images are from real life any way. LiberalArkie Jun 2025 #12
Combine this with Palantir's data compilation misanthrope Jun 2025 #14
Wouldn't you be able to find out how the video was made, like what kind of camera Beringia Jun 2025 #15
"I've always loved Big Brother." Kid Berwyn Jun 2025 #16

Johnny2X2X

(23,685 posts)
3. Fingers
Mon Jun 2, 2025, 09:17 AM
Jun 2025

That's the only way to tell what is AI right now, AI still messes up fingers. Sometimes it's extra fingers, more often now it's just longer fingers, like a thumb that is longer than the rest of the fingers.

Once AI can do fingers correctly, there will be no way for me to tell anymore.

0rganism

(25,472 posts)
13. Already addressed. IOW, there's no way for you to tell anymore
Wed Jun 4, 2025, 05:05 PM
Jun 2025

AI image and video generators have largely resolved the issues with extra/missing/malformed fingers and teeth. They're getting better at rendering occlusal shadows and reflections too. If you see any images or videos with that kind of inconsistency in them now, it's mainly because the prompter didn't care about fooling people anyway.

We have a serious problem. It's here, now. So far there's not a plan to deal with it.

hunter

(40,368 posts)
7. Okay, who's my stalker?
Mon Jun 2, 2025, 09:33 AM
Jun 2025

""Oh my lord, look at that Atari 800 you have behind you! I can't believe how nice it is!"


 

k_buddy762

(638 posts)
8. It can't do text very well
Mon Jun 2, 2025, 10:30 AM
Jun 2025

and everything looks cinematic.

In year, yes, you will be unable to tell reality from fiction.

Buns_of_Fire

(19,007 posts)
9. Roger Stone is already limbering up his tentacles for future political ads.
Mon Jun 2, 2025, 11:12 AM
Jun 2025

Hmmm... Roger Stone as Cthulhu... I see some possibilities there...

TheProle

(3,900 posts)
10. It's impressive technology
Mon Jun 2, 2025, 12:31 PM
Jun 2025

and I bet there are a lot of film and commercial directors getting very nervous.

Abolishinist

(2,887 posts)
11. Wow, amazing! Thanks for posting this.
Wed Jun 4, 2025, 04:40 PM
Jun 2025

I'm curious, would anyone happen to know the answer to the following. If one creates a video, such as the one I linked, are they able to replicate the person and have them appear in another video, or are they one-offs?

https://cdn.arstechnica.net/wp-content/uploads/2025/05/A_man_on_2025052615221.mp4?_=38


misanthrope

(9,380 posts)
14. Combine this with Palantir's data compilation
Wed Jun 4, 2025, 06:17 PM
Jun 2025

It would become very easy to frame someone for a crime, regardless of what they had actually done or not done.

Beringia

(5,343 posts)
15. Wouldn't you be able to find out how the video was made, like what kind of camera
Wed Jun 4, 2025, 07:04 PM
Jun 2025

You can't fake things like models of cameras and ownership of the camera

Latest Discussions»General Discussion»Holy crap: AI video ju...