Gemini Omni
deepmind.google230 points by meetpateltech 6 hours ago
230 points by meetpateltech 6 hours ago
In my day job I program rigid body behaviour in real time amongst other simulations. I think rigid body contact is hard to learn as it is inherently discontinuous.. something you discover when trying to code a solver.
As such I always use this prompt as a test: "A video of a jenga brick tower falling over as a brick is removed. The physics of each brick must be realistic."
It gave me a video of where bricks suddenly disapper or morph into others[1]. The linked video is after 2-3 iterations of me insisting on realistic physics. If you are just glancing at this, you would believe it is realistic.
That said this is still very impressive and one more step towards .. IDK what. But I am a bit reasurred that at least my job won't be fully replaced with AI :)
> But I am a bit reasurred that at least my job won't be fully replaced with AI :)
I honestly can't comment with certainty that training from videos alone and whatever tokenization scheme they're using will ever get perfect dynamics.
However it is worth noting that transformers can do a pretty good job at learning dynamics with the right pipeline (not video): https://arxiv.org/pdf/2605.15305 https://arxiv.org/pdf/2605.09196
My point here being that representationally, it might be possible to learn good dynamics without a radically different approach/arch. There are already models that extract 3D tracking points from videos, so they could possibly be leveraged for learning dynamics (which on its own gives precedent for end-to-end approaches also possibly working).
Thanks for the additional reading. I've often thought about LLMs and their ability to represent the physical world with its laws. And always concluded it is not really possible to do so with "just" text tokens and their relations in a latent space. It looks to me there are different approaches being taken to tackle this:
* You could instruct your LLM to interact with an simulator to run experiments and infer behaviour
* You could edit the transformer model and inject spatially relevant data rather than text as is done in above paper
* You could change the architecture to be more condusive for representating a world state. I.e., LeCun's JEPA world model.
* You could further enhance some of the above by using a differentiable physics engine (eg. NVIDIA Newton) to calculate losses directly.
But at the end of the day if a model has any hope to always produce realistic physics, it HAS to learn the laws of nature in some form or other. It looks to me that the next big leap could be achieved by combining the last two approaches.
P.S.: I like discussing such topics. If anyone knows a forum or discord with like-minded people, please let me know :)
Totally unrelated, but what would you say the feasibility of writing simulation software for simulation of/replicating body movements during/in a martial arts technique would be?
I’ve often thought it would be very handy to have a proper simulator for being able to simulate and identify inefficiencies in one’s technique, but no idea whether it would be feasible to do.
Would be similar to the typical simulations of humanoids. If you need to model the deformations of the human body, or get a proper model of tendons that make up humans, it'll be more difficult, but possible.
Proper simulators for those exist, you essentially need an engine with a compliant contact model. MuJoCo is the goto here, see:
https://mujoco.readthedocs.io/en/stable/modeling.html#muscle... https://mujoco.readthedocs.io/en/stable/computation/fluid.ht...
These explicitly model biological muscles. IIRC it was originally created to model human hands (I could be misremembering though).
Really depends on the fidelity you want.
Edit: I also work in rigid body simulation for robotics.
Such videos are essentially dreams: how it feels that the planks should move, not what equations of rigid body physics would compute. And the feeling is realistic (even if overly dramatic in the end). If "stylistic transfer" works for static pictures spread out in space, why won't it work for the character of motion spread out in time?
I wonder what's the training data that makes it generate the final "explosion"...
thanks for intro to streamable
In my experience (from a couple of years ago), Streamable can be great but it's just worth checking what their current retention policy is like.
We were sharing game clips with each other and after a while realised our old clips were just gone, being deleted after 30 or 90 days or something.
At first usage I'm not impressed. I've probably spent a couple grand on Seedance 2 to date, and I can't find anything google omni flash does better than Seedance from running a handful of samples through the system. You can find some of the videos I've made in my HN bio link.
Just curious - are you at all concerned about the legal implications of ai-generating property listing videos?
The legal risk probably lies solely with those who are selling the properties. They are responsible if the video misrepresents anything.
yeah, it's all about keeping everything grounded in reality.
But it looks so fake that I wouldn't waste time visiting a property advertised like that.
I have exactly the same thought. Anyone who had used seedance 2.0 a bit can tell Gemini is a bit behind, and seedance 2.1 is on the horizontal already.
Seedance 2 is amazing, compared with anything else American tech is producing. It does struggle with consistency like all other models.
The other problem is Seedance is heavily censored because of copyright concerns.
> The other problem is Seedance is heavily censored because of copyright concerns.
While at a cursory glance it looks as impressive as always, subtle spatial errors, and geometry that changes as it goes out of sight and comes back again hints at the fact that Google has still yet to solve the problem of deep spatial understanding.
Which considering just how pretty and detailed this whole thing looks, imo points at a fundamental issue at how these things are trained - it's as if there's no structure to its knowledge and training, like how an artist trained to draw would first try to understand simple 2d composition, then perspective, then light and shadow, mastering each concept and gradually building up a hierarchical understanding - it seems like its trying to learn everything at once.
I would rather see an AI model that I could give a floorplan of a building and it would generate an accurate flythrough on any path, even if it looked like butt.
Im not just talking out of my arse, I did work for a while in data science/engineering, and one of the big lessons people needed to be reminded of is to clean/downsample the data - a dataset consisting of a million samples could very well take 1000x as long to process as if we downsampled the whole thing to just a couple of thousand samples and we could learn the same conclusions with the fraction of expended time/effort.
I'm sure there's a similar logic in RL, that if you dump a trillion samples into the datacenter that consumes the same power as a city, what the model learns is what it could've learned with a much more curated training set and directed approaches.
> Prompt: Make it look like the weird shape of my hand hole super zooms and magnifies the ground it's looking at in sharper quality.
There's got to be a reason this is phrased so insanely, right?
Even weirder:
> Prompt: A skeuomorphism stop motion explainer about how the brain hippocampus works with a compelling voiceover. Don’t add seahorses. No voice cuts at the end. Don’t add text
Seahorses???
The genus of the seahorse is _Hippocampus_.
And the fact that a transformer model can't distinguish between the two in the context of the sentence given is a point against the general nature of the intelligence.
We are at the point where "don't add seahorses" doesn't actually fill it with seahorses like the previous models.
Yes, if you watch the video closely you can see that the "lensing" effect only really covers a circular area—this prompt probably went through multiple iterations where the author was trying to improve it so that the shape of the hand was reflected more closely.
At the bottom there is a "Try in Youtube Shorts" button.
Oh god...
I mean if we're just blasting past our climate tipping points anyhow, why not just actively dump entire lakes' worth of water out for people to post slop for clout, right?
May as well power off the whole grid now and have the Amish start teaching us how to survive
We could be solving fusion power and instead we’re generating videos of birds in space or something. The market is a harsh mistress sometimes.
I'm an AI optimist. But AI video is probably the one thing that does depress me. Seeing that we can make anything visually, there's nothing that impresses me visually. I watch a video that two years ago I would've thought was really cool, and now my first thought is, "Yawn, is this AI?".
Video, more than anything else, is the place where I really care if something is AI or not. If I could get a TikTok that had no AI usage -- I'd be in. Which is weird for me, because I'm typically the guy who is all-in on AI.
It ruined the whole category of "cute animals acting goofy" content for sure.
Yeah, I'm kinda sad about that one. Most of my friends and family are aware many of these are fake now, but argue that it still invokes the same response in us so it's okay. For me, though, however intangible or irrational it may be, I do feel a sense of loss.
Funny enough, this is actually one of the few things which has bothered me with the AI boom, and I'm mostly pro-acceleration. A lot of what's happening seems inevitable. But surprisingly, knowing that cat or dog or bird or lizard or butterfly or whatever has a strong chance of being generated really does take something out of it to my mind. And I say that also knowing the extreme amount of staging which has long gone on with traditional nature videography. Somehow, knowing the animal is real means something... I'm still trying to figure out how to better understand and express this.
You get back as much as you put in. Just like with all generative tools the quality of the output depends on the quality of input. Slapping a prompt together will only get you so far, if you want the models to generate something really striking and unique you need to get your hands dirty. Gotta break out ComfyUI and build yourself a specific workflow, once you dig deep and understand how things are put together, why and so on, you can make really amazing stuff with any generative models. But you have to pay for that experience in patience and knowledge.
For a few weeks, YouTube thought I wanted to see videos of package thieves being surprised by a booby-trapped box that was actually a glitter bomb. Video after video were these AI created shorts of supposed doorbell camera footage showing a thief running away with a box that explodes into a giant pink cloud.
I eventually picked one and opened the comments and the top comment was something like "This is obviously an AI video. Who watches this?" and the reply was along the lines of "me because I like seeing thieves get what's coming to them".
So you, like me, aren't interested in AI videos but I think there's a lot of people who don't care if it's real or not.
Thankfully, YouTube eventually stopped showing those to me. Now it thinks I'm interested in road rage videos. My YouTube feed outside of the three of four channels I've subscribed to is terrible.
> and the reply was along the lines of "me because I like seeing thieves get what's coming to them".
I really wish a subject matter expert would pitch in to tell us what this is about?
like a totally made up thing that is fake, somehow gives a sense of justice and satisfaction?
is it something about imagining it happening in reality, or what?
for me, if I see that something is AI, it's like I just feel nothing. because there's nothing in it, it has nothing of real value? like it doesn't evoke anything in me, it doesn't make me think "this was a great find!" or make me want to send a link over to my friends, etc.
This is the whole Dhar Mann genre, which is so cringe, but it definitely tickles something in us.
Do you ever feel a sense of satisfaction watching a movie? I'm thinking of scenarios like when the bad guy is finally defeated or the hero achieves their goal.
I think the opposite. It allows more people to be creative. Similar to how the DAW allowed more people to become musicians. You can produce a hit song with just a laptop now.
Now you can have people producing videos without needing a crew of people.
You never needed a crew of people to make videos. This is just outsourcing people's creativity.
The potential for harm is so much greater with video than creating an mp3. You can stoke hate and fear so easily.
The method in the madness is to generate so much on demand slop no one will accidentally find your hate and FUD content anyway.
It's not all bad: https://www.tiktok.com/@openchub/video/7641631412407274782
I tried to watch it, but TikTok kept throwing up a dialog over top asking me to slide a puzzle piece into place. I did three or four before just closing it.
Browser crashes while scrolling because of all the auto playing videos. Please use IntersectionObserver to pause the video when not in display.
Not to negate your experience, but seems fine on Firefox 150 on my Windows ThinkPad X1.
On my iPad Pro from 2017, none of the videos even play. Not sure what's better!
blog post: https://blog.google/innovation-and-ai/models-and-research/ge...
model card: https://deepmind.google/models/model-cards/gemini-omni-flash...
What I'm hoping/waiting for is IMDB users creating alternative endings of movies.
It could make the comments section even more fun.
> I can create more videos as soon as your limit resets. Check your usage in Settings
I did not create any videos yet.
Google, building great AI that nobody can try out.
But thx for the press release.
I think Hollywood is in for a rough era. The disruption is happening at break neck speeds.
At one point the only way to know if something is real or by a major US tech company is nudity.
Hollywood is already in a rough era but it’s because they can’t create original human stories any more.
This tech won’t change anything.
Yeah, during most blockbuster movies lately all I can think is: "All pixels, no plot."
Project Hail Mary was largely real sets and a puppet.
They used a puppet to play Rocky? Was not sure how they did it, don’t think I would really care, but that’s cool.
They used a puppet on-set for in-camera reference but most of what you saw on-screen in most shots was CGI. I wrote more here: https://news.ycombinator.com/item?id=48198851
Was it? Hollywood has been caught lying about that to seem more authentic before. https://www.redsharknews.com/why-do-movies-still-insist-ther...
There are tons of behind the scenes pictures and video of the Rocky puppet being used on set, and Andy Weir talks in interviews about how almost no CG was used to enhance the puppet. I guess it's possible to fake all that, but it's a lot of lie to cover up.
Andy Weir is a wonderful novelist and was truthfully relating his understanding but he's not a VFX person.
I didn't see the quote you did but he probably confused the fact that PHM used physical elements in place of some CGI in certain scenes and the separate fact that a realistic physical puppet was used on set for reference. Some parts of that puppet are seen on-screen in some shots but most of the creature in most shots was CGI or CG enhanced (which looked great thanks to the ideal in-camera puppet reference it replaced). I explained more here: https://news.ycombinator.com/item?id=48198851
I agree PHM was great (and I loved the book before the movie). But as a VFX person, please be careful not to buy into the currently popular studio PR line: "it's all real, almost no CGI". Media and influencers love this line and often unknowingly muddle the studio's very carefully crafted press release wording into outright lies by paraphrasing and making assumptions. The problem is these aren't just white lies, they deprive some very talented VFX artists from getting credit for amazing work.
About the misunderstood puppet: A real Rocky puppet was indeed used on set (actually a few different puppets) and some of the puppet is sometimes seen on camera. But most of the puppet was digitally replaced with CGI or CGI-enhanced in most of the scenes. However, using a much more realistic puppet on set is indeed notable but not because the character wasn't CGI. The puppet is worth talking about because it directly enabled the final mostly-CGI character be really good CGI. It's good because shooting the physical puppet gave the VFX character animators an ideal reference that's "grounded" in the physical reality of the set, camera and lens. The subtle interplay of light, shadow, texture and specularity in the CGI are all grounded in reality. The puppet also let the actor interact with something closer to reality. It's a wonderful technique and should be celebrated instead of obfuscated to promote a "No CGI!" falsehood that trends well on social media.
Also, PHM did use real sets (like most movies) and they were able to avoid using green screen for some of the ship exteriors but those backgrounds were still digitally replaced with CGI rendered elements, they just didn't use green screen to pull the matte. But on social media, "No green screen" (true) was conflated into "No CGI" (false). Instead of green screen they used a black backdrop with careful lighting and some hand rotoscoping to extract the digital mattes. Doing it this way had the advantage of not needing to digitally remove green spill on reflective surfaces by hand and it saved money over doing a StageCraft virtual volume at that size. Done well, a green screen could have produced the exact same shot but it would have cost more and taken longer.
But influencers and media are unintentionally perpetuating "No CGI" myths instead of focusing on the actually interesting, more nuanced reality. Using more and better physically grounded references on-set IS a breakthrough that helps turn bad CGI into great CGI. Another example is Top Gun where "artfully misleading but technically true wording" in studio press releases grew into outright falsehoods online. Tom Cruise was truthful in saying that he was flown in a jet right alongside other REAL jets doing simulated dog-fighting. The lost nuance is that all the other jets Cruise flew with in those dog fight scenes were old Soviet trainer jets that look quite different and are much smaller than real MIGs. So the trainer jets were entirely replaced by CGI MIGs in post and are never seen in the final film. And we couldn't tell because the digitally removed jets provided ideal grounded reference for the CGI pixels that replaced them. And that's how we ended up with several famous YouTubers proclaiming "These are REAL jets, not CGI!" while showing 100% CGI jets. Same with Wicked and the CGI tulips. The fact that Wicked used thousands of specially grown tulips on-set (true) was confused into proclaiming "ALL these tulips are real, no CGI!" (false) while showing a scene where >90% of the tulips were CGI.
you would watch a movie generated with the sterility of an LLM?
AI is already in a bunch of creative workflows. Just look at modern Photoshop. Selecting and hitting delete has AI infill for the background replacement.
Creates can these video gen AI in various ways. There are some youtube channels of people using these in creative workflows that are really impressive, from mocap replacement, character insertion, background replacement, changing camera angle in post, animating/inserting characters from character boards, animated between stills generated in traditional methods, etc. It's not just "prompt and generate". It can be, because it's easy, but it also doesn't have to be. It's a tool.
i do photo restoration as part of my research (bizarre place to be for a math undergrad), so i do think AI is a lifesaver for very small adjustments that would be tedious or subpar otherwise. i just disagree that its creative output is of value (which isn't the case you made, anyway).
I do wonder how studios are working around consistent human faces, it's a problem on almost every discussion forum I have read for AI videos and not something that seems to be solved yet.
Do you have any examples of those creative workflows that have made it into Hollywood for example?
Weirdly phrased, but yes, I would watch a movie generated with an LLM by a person passionate about the movie they're creating.
I think Hollywood's obsession with unnecessary sex scenes[0] is the #1 reason I have been watching less and less movies. So yeah, probably.
[0] e.g. Don't Look Up
Sure; why not? It has to be better than some of the absolute garbage that's out on the various streaming services today; right?
god help us if we have to choose between the two );
I’m willing to condition long duration copyright on streamers being able to implement mature content edits.
Me? No. My kids? I think they already have. I don’t allow YouTube in our house, but they for sure watch slop with friends.
At the moment the duration of each shot is a major limitation. When that limitation gets solved is when we'll see actual disruption.
Average shot length is down to something like 3 seconds in modern cinema. That's a pretty low bar.
Even though I don't have words to express how impressive this capability looks. I am genuinely scared at the harmful use cases of this.
So it's really good, and we have reason to believe, never again, anything that happens in a video. Unless there's a super-product somewhere to authenticate footage?
Now that they've broken the ability to trust video, they're looking to build it back, as long as you're allowed to use the tools:
https://blog.google/innovation-and-ai/products/identifying-a...
(and the previous SynthID: https://deepmind.google/blog/identifying-ai-generated-images...)
But it very much is "close the barn door after the horse has bolted and the barn has otherwise burned down".
It seems like this super-product will have to be a thing soon or we will have to just stop using video evidence in court and other critical applications
Who is creative enough to drive this in any meaningful way?
Certainly not me - you have to be a great artist /designer to even imagine what to do with it.
Back in 90s during the first wave of the desktop video revolution when desktop editing became possible and consumer camcorders got pretty good, there was a popular marketing slogan: "Now your imagination is the only limit."
I used to joke that was the moment we discovered "for most people that's a pretty big limit."
Does anyone else feel like Google is just always a dollar short and a day late here? Maybe not a dollar short, but it's like they've consistently been focused on the wrong thing. First they missed chatbots, now they're missing coding agents while they double down on chatbots and video gen (which OpenAI has already basically abandoned). Maybe this strategy is actually genius and I'm too stupid to grasp it.