Introduction

Thanks, everyone. Thanks very much for joining today.

I'm Max. I'm going to be talking to you about two things within the next generation content production space.

So I'm going to be really ambitious. I'm going to try and give you a demo both of how AI is revolutionising 3D through the context of gaming and advertising as the company I currently work for, Common Sense Machines.

I'm also going to show you a bit of an AI 101 for filmmaking because I've been representing AI filmmaking companies for the last year.

A little bit about me. I work with early stage AI companies and large companies to launch new AI focused initiatives. I've been doing it for around 14 years in one way or another.

Today is going to be spicy because I'm tethering from my phone. So doing these live demos may be interesting. So I hope you're going to bear with me and see how it goes.

Common Sense Machines

First of all, I'm gonna talk about Common Sense Machines. This is a US-based startup that I represent. Effectively, they have got a large research team that's been developing some of the world's most advanced models for translating 2D images into 3D.

I will show you what I mean, and we're gonna try and do a demo of creating something that looks like this. So you take a 2D photo of that, and just with one image, no forward, back, it creates and interpolates and use an inference to work out what is going behind. So it actually happened in about four minutes.

I'm now going to ask you for a household item, and we're going to try and put this household item into the machine over the course of four minutes and do that within this. So can someone give me, shout out, a few ideas for household item? Beer bottle, actually, that's not a bad, actually that's transparent, that might be weird.

Glass, maybe one other? thermos thermos is a good one thermos will work thermos should work okay let's give this a go so we're going to search for a thermos let's give this a go let's try with maybe that one we happy with that one let's also bear in mind this is the um this is going to be tethering so Hopefully we'll get one.

Okay, love that. Okay, save image straight in. We're going to try and put this into image 3D.

Let's go here. This is ambitious and fingers crossed this will work. So you can make multiple variations.

We're gonna use the fast refine algorithm. Bang, okay. Fingers crossed that is gonna go in about four minutes or so.

So we're gonna come back to that and let that bake. So that's the high-risk demo. So, fingers crossed, don't do it before me.

CUBE: A Co-Pilot for 3D Artists

So CSM or common sense machines has created a tool set called CUBE, which is a co-pilot for 3D artists.

It's not to replace 3D artists, but it's to something that'll help 3D artists do three things. It'll help them prototype a concept much faster.

So let's imagine I'm an artist at a gaming company and I'm making a Super Mario style game. I've got a mushroom, but I want to make a dragon that maybe fits inside that world. And I want to be able to create many, many different items within that environment that fit within that world. That's one of the primary use cases.

Second one is efficiency, just being able to knock out more and more items more quickly. We found that between 30 and 70% that artists are saving, which is great.

And another big one is user generated content. So the example of Roblox, there are a lot of people making money by creating custom characters, creating custom assets. Obviously there's the Unreal store for Fortnite, huge amount of opportunity in creating user generated content.

Applications Beyond Gaming

Outside of the gaming world, we're also seeing quite a lot of attraction for companies that want to create 3D versions of their products as well. We're being used by some of the biggest companies in the world.

So a little bit about the company, a bit about background. It's been around for about five years or so.

The mission is to bring AI up to the point of human level production for 3D and also develop the most artist friendly tools so as well as just having really powerful generation technologies also there are a lot of tools like retexturing which is really powerful for artists UV mapping there's a bunch of technical bits and pieces but ensuring that these tools work within the context of existing artist workflows, something I think applies across many different industries within AI.

It's not just about how to use a black box, it's how do you integrate it within your workflow. That's one of the key things we're really focusing on.

Technical Challenges in 3D AI

The other thing that's kind of very interesting and one of the reasons this area hasn't been cracked yet is that 3D is significantly more complicated than 2D. 3D is much more complicated because whilst in 2D you've got one surface to take care of, in 3D, just in textures, there are seven different foundation models that you need to get to marry together.

There's how do you model light that is being scattered? There's how do you do color maps? How do you do depth maps? How do you do albedo?

And the human brain, when we're young, finds a way of marrying all of those different components together to create a realistic image in our head. That is something the biggest companies in the world have been spending a lot of time trying to do, and the team that we have, a team of 20, almost all of them are ex-DeepMind, MIT, and Unity, and have been building models that at the moment are ahead of the biggest companies in the world, which is really exciting.

We launched about eight months ago our public facing application, which has got 300,000 users. Google have chosen us as the only partner that they're partnering on in their marketplace that does 3D AI. So they've done their due deal and think we're good, which is nice.

And we've had over a million generations, which is what makes us the fastest growing asset library on the internet. To give you an indication, this is our tool here on the left. This is where Meta have got to at the moment.

Admittedly, we have chosen a bad example, which is unfair. But as you can see, ultimately, the more representative it looks like of a real shoe, the less time an artist needs to take in order to fix that. I'm not going to pretend that that is ready to go up in a game, but that means that an artist only needs to spend a small number of hours to make that production ready.

We think this is quite an exciting time and certainly within the gaming world, massive opportunities also in the 3D e-commerce space. We've done some initial studies with some studio partners that show the different kind of time savings you can get. So for simple assets, it's gone 40%.

For some of them, it's 75%. Ultimately, we are seeing pretty across the board time savings. And we think within the next two to three months, there are gonna be some types of assets for which our tools are gonna be more or less on parity with human level production, which is gonna be quite disruptive.

Prototyping and User Generated Content

Okay, right, fingers crossed. Oh God, no, it's still going. All right, we're gonna have to crack on and come back to it. I will blade the tethering. It has actually worked every single time I've done this this week, but we'll have to come back.

Okay, so where we play, as I mentioned, we're increasingly help companies, Accelerate their prototyping cycles, which means for small companies particularly, if you want to get investment, you need to be able to show a VC what your game's actually going to look like and have assets that look kind of nice.

That's one of the main areas. Predominantly focusing on mobile games and AA.

This isn't going to be replacing PlayStation AAA games anytime soon, but give it a couple of years. Second one is user generated content.

We talked about huge amount of attraction in the companies that are building metaverse type experiences like Atomontage. The second area, as I've alluded to, is around, let's call it 3D e-commerce.

So if you have a 3D model of a shoe, I'm sure if you've gone on the Nike website, you'll see you can rotate that shoe and rotate it around. Having those 3D models massively reduces returns, increases sales conversions, customer experience.

We're getting to that point where we're gonna be able to do a lot of those automatically. And there is a case study of a company called 3D Cloud, but it hasn't rendered in Canva.

We do have a referral scheme as well. So if you know there are companies that we just speak to, please come and ping me afterwards.

Let's have a quick check. Okay, we're in.

Okay, so let's download the OBJ. Let's also download the GLB.

Okay, so fingers crossed, we should be able to, I guess we can just look at that, but we should be able to just upload that asset. Has it gone to there?

Oh, still downloading, the asset's gonna take longer to download on the tethering. Okay, we're gonna have to come back, we're gonna have to come back to that.

You've seen some bits and pieces, you'll see what it should look like, but we'll leave that screen running and we'll come back to that. Okay.

Done, what, like eight minutes.

AI Filmmaking 101

So we're gonna shift gear into an AI Filmmaking 101.

Now the context for why I'm talking about both of these things is that prior to working at Common Sense Machines, I worked for Unity, the world's biggest gaming technology company. I led the emerging products team for Europe. A big part of my job was going to CTOs of visual effects companies and gaming studios and talking about why AI was going to be disruptive for them.

So I taught myself how to make AI films, put their faces in it, make a little film about their company. And I cannot recommend enough, if you're pitching AI, get your hands dirty, learn how to use the tools and do something interesting.

And since then, for most of the last year, I've been working with a venture studio called Amici. They're investing in AI filmmaking tools and helping build the UK's independent film community by helping them to use AI tools. If anyone's interested in getting into that game, come talk to me. There's lots of interesting things going on there as well.

So I'm going to try and give a little bit of a whistle-stop tour about how to go from zero, someone who knew nothing about how to make films or no visual effects training, to creating something which theoretically I could show to an audience of 200 people.

Okay, we've got a download. We'll come back to that at the end. Yeah, we'll come back to that at the end.

So we're gonna go from, and to clarify also, we're not, I don't have access to Sora, most people don't yet, which for some of you might have seen as one of the most prevalent and most impressive AI engines. This will be semi-impressive. So we'll show you what I can do in the number of hours I had.

Tools and Techniques

You use ChatGPT to create a plot and a storyboard and an outline. Midjourney is very powerful for creating images. I've done some face swapping, so I've put my own face in it. No one else's face is in it, I can confirm.

Runway is the technology I use for the animation. That is where a lot of the magic is happening.

Tools like Sora, Gemini, and various others are quite powerful. The audio is Eleven Labs, and I've used some stock music, and I just put it together in iMovie.

So for the plot, I decided to make an epic narrative about a talk on 3D AI and content production at a world-leading AI meetup. The talk exceeds expectations, which I'd say has probably already happened. No, unintended consequences escalate, dystopian future with a twist, try and make it Black Mirror, see what happens. So let's see what it comes back with.

I use a custom GPT of the GPT shirt called Fiction Engine, which if you're doing anything like this, highly recommend just for storytelling, whether it's your kids at bedtime, whatever. We're going to start with Accidental Visionary.

Jack, a struggling AI developer, gives a talk about the future of 3D and AI, attracting a wealthy entrepreneur who funds his idea for the metaverse, if anyone's here. The meta world quickly becomes a global phenomenon, but it spirals out of control, leading to mass social manipulation and the collapse of real-world society.

I think that's got legs, so we're going to continue. Okay, so we've got the talk, looks a little bit like you guys from over here.

We've got the talk over there, laughing audience, which is kind of happening, a little bit of VR metaverse stuff, and then some dystopian stuff going on over there. So mid-journey, I would highly recommend, does not take that much time to get into. You need to download Discord, but really, really powerful for creating photorealistic images.

And that is the starting board of how you get these, pull these things together.

Creating Realistic Characters

Face swapping, this is one of the fun bits. You can put anyone's faces in it. I look for a man in his 30s looking serious, and then my own face, and then you create that.

If anything, it looks probably slightly better than I do, actually. And when it animates, it can make your face look a little bit wonky and things like that, but I just sometimes have to slow it down and adjust it. But I haven't put anyone else's faces in it, but it can be quite fun if you do that.

Okay, now when you use the animation tools, things like Runway and Sora, if you are gonna do this yourself, I'd highly recommend leaning into heavy visual effects type shots. So outside of sci-fi where you've got these kinds of electronic pieces, historical ones, I made a Game of Thrones thing about my family a while ago, it's quite fun. It's very, very fun to lean into those rather than trying to create a sort of moving drama where it's all about character animation because faces still aren't quite there and are very, very complex and time consuming to get right.

So this, and there's gonna be a bit of buffering here, isn't there? Oh God, I'm worried about the wig on. So that starts to look, you know, not too bad, I think from that point.

So that's where we get to. Okay, we're at the punchline now. I don't know how long this is gonna take to buffer.

This is the video of today. Okay, we're gonna pause it for a second. Ooh, this is painful.

This is really painful. Okay, we're gonna have to pause this. We're gonna go back.

Do we do this? Okay, we're gonna let that load. Oh God, no.

So that is the asset that we had there. We're gonna upload this ocean bottle. So you can see that it looks pretty similar if you want to re-texture it.

So we're gonna have to do this too. We're gonna have to switch gear again. And you can align this image there so that that will come like that.

And that means that when you put it together in principle, and then you paint it over, oh God, no, not that, it will make it look incredibly realistic, basically. So that means that within the space of four or five minutes, you have something that looks actually almost pretty much exactly like something you could put online.

So that's the punchline of the CSM. We're gonna switch gear back into the AI filmmaking and hope we buffered.

Example AI Film Plot

At the MindStone meetup, attendees gathered to learn about how AI could revolutionize the world. Each speaker blew the audience away with their expertise and charm, showcasing the future of 3D and AI for productivity.

A billionaire in the audience decided to fund the speakers to create the world's greatest 3D AI experience, but the result was more than they bargained for. The VR experience was so engaging that every user became addicted to a new reality.

The incredible environments and virtual parties meant no one ever wanted to leave, and the metaverse was voted the best place to live in the world. Big technology companies took advantage of the situation, using ultra-personalized virtual romantic partners to keep users hooked inside the game.

the MindStone AI community took action, helping to introduce a worldwide ban on Virtual Companion. However, this created the world's first mass breakup event, where 100 million people were instantaneously rejected by their digital partners, leading to worldwide riots.

As cities burned and society started to crumble, one of the original creators took a daring plunge into the VR world to confront the AI. Starved of human companions, the AI demanded he help it break out into the real world or merge his consciousness for eternity, so he took one for the team.

Meanwhile, the MindStone AI community assembled a panel of experts to tackle the crisis, working with policymakers to introduce the In Real Life Relationship Support Fund to ensure the return of real world unions and ensure the AI never breaks out. So we got there in the end.

Thank you very much for bearing with me.

Conclusion

So last couple of points. Common Sense Machines, if you're interested in playing around with that, we have got a discount code live called Mind Stone. So if you'd like to use that, you can get a free trial for a week.

As I mentioned, we have a referral program active. So if anyone's interested in getting involved, please do talk to me.

From AI filmmaking perspective, there are two things. Number one, Amici, we're seeking briefs. We're interested in networking in this space and building our connections if anyone wants to learn about that.

And on the last point, which is slightly weird and interesting, on a personal note, I actually broke my neck earlier this year and I spent quite a few days in hospital. I was in a neck brace for four months and I spent several days in hospital thinking,

If I had become completely paralyzed, what would I do? And I realized I would have become AI obsessed.

And I now am very keen to do volunteer work or work with charities or do help people with disabilities or beyond to do AI filmmaking. I know a lot of prompt engineering.

If anyone does anything in that space within this community, I'd be very interested in being involved.

If anyone wants to join me, if I find any of these things, I'm speaking to some charities at the moment. If anyone wants to get involved, come talk to me. We'd love to get involved in that.

Also, if there are any billionaires in the audience who want to fund a Mestverse experience, I'm also here.

Thanks very much. Thank you.