Stay Curious: Jerrod Lew on AI Video

The main message I want to give to everyone is this is a time where I really recommend you stay curious and stay active with all the changes and updates that are happening in the AI space, not just with art and video, but in general. I think this is a really, really big year where we're going to see some major launches in the creative space, but also in the personal AI space as well. So I encourage curiosity this year as much as possible, and don't be afraid to dive in and try new things.

Welcome to the Imagine AI Live podcast. I'm your host, Chris Madden. And today we have with us Jared Liu. Jared is an AI creator, and he's deep in the space of AI video creation.

He's worked with some big names, including Adobe, Runway, and FreePick. And now he spends his time speaking on AI, consulting in AI, and educating through his personal channels. Jared, how are you doing? Doing really well.

So happy to be here. Imagine AI Live is amazing. We've obviously had a few events in the past, and it's exciting to see these kind of conversations come back because they let people talk about things they're passionate about, what's happening in the space, and I think people see how broad the AI space is in general. And so happy to have opportunity to come back on.

Definitely. So I want to make sure that we get into like your favorite tools you're using right now and everything. But first, I want to talk about Sora. And OpenAI decided to shut down Sora.

And I know that you played around a lot with that you were almost maybe a power user of the platform, creating things being creative and using the app. Why did Sora shut down? I wasn't really fully surprised when it shut down. I think I was more surprises at the abruptness of it because it looked like all the cards were being played perfectly for open ai they had the disney deal they launched an app they had a genuinely good video model that had a lot of hype beforehand and they really capitalized on it and people look at the funny videos of the soar app and stuff like this but you don't realize as well it was actually genuinely a very powerful video model it was great i just think the aim of open ai has changed But not just them, but amongst other tools out there as well.

When these tools opened up, they were really built for consumers. And I just think it costs a lot of money to run projects like this and create and maintain models to the point where I felt there was other places where they could get more revenue. They ran adverts, Super Bowl ads and stuff like this. And all of it was focused on coding and tried GPT codecs and all these things rather than the creative standpoint.

So I just think there was a decision made up above that said, look, it just wasn't making enough money for them to maintain it. I think there will always be some sort of creative element in chat GPT and OpenAI. Obviously, they have their own GPT models for image, but I don't see them going back down the video. But it's really unfortunate, honestly, because it was generally a good model.

And the way they just kind of switched it off abruptly, it was quite harsh. They just had to pivot. Did you have a favorite video that comes to mind that you made on Sora? I had so many funny ones of me, like a video of a security officer fighting me, stealing fish from a store and running out.

And just the whole hilarity of how it would cut the video so seamlessly to make it funny. It was fantastic as a time, the time to do it. And I really felt that I could be something that kicked off because it was personal. It was us, ourselves in the video, all from a single image.

You could tag celebrities to be in the videos with you. We could make our own IPs and stories based on us, but it just didn't really click for whatever reason. But all the early videos I made to start were absolutely hilarious for me. Yeah, I made a few myself and we need to go back in there and make sure to download them all for posterity's sake.

So moving into the tools that you're using now, what would you say are the three most powerful AI video generation tools, software out there on the market today? I think the most powerful one right now is C-Dance. So model C-Dance 2.0, that launched or was announced a little while ago, and then some people got access to it a bit earlier.

This is by far the most powerful model I've ever seen. The fact that you can prompt with text, images, audio, but also a video means that you can extend videos as long as you want. and it keeps the context of the previous video it's extending from. The realism is fantastic.

There's dialogue, audio, sound effects, all this stuff that's embedded in as a norm. And now they just announced that you can render up to 1080p, which means you can get really production-level stuff right out of the box without needing to upscale it with a third-party tool. So it's an extremely powerful model and it has other features like editing. If you want to edit a video, you can upload a video you've generated and change the scene, remove things, add things, modify weather, environment, characters.

You can obviously prompt for different music and dialogue and sound effects as you want as well. All these things combined make it the perfect tool right now because it has the quality, but all the features that back that up. It was pretty mind-blowing getting access to this. The way it cuts video, it's a way you can prompt each individual shot whilst keeping the environment and the character intact.

It's amazing. If this is the standard of how we're starting this year out, I'm curious to see some of the stories that we're going to see people tell with this model this year. So that's the number one outstanding model right now. Diving a little deeper in there, Seedance, who runs that one?

And give our viewers an idea of how much it costs for a subscription to use it. It's a model that's run by ByteDance. They own TikTok and a bunch of other things. The cool thing now with Seedance is that it's now available across multi-platforms.

So even if you don want to subscribe to CapCut to use it there or Jemina to use it there which are ByteDance companies You can now access it on other tools like FreePick and Runway that are open to third models coming in And that changes a lot for the consumer because it means you get more for subscriptions for those tools. If you already have subscriptions to those platforms, you probably already have access to CDance 2.0. And I actually recommend that as a way to get access to it.

All right. Well, so that's your most powerful AI video platform software out there right now. What's your runner up? Runner up is Kling.

So Kling launched 3.0. They're another Asia-based company. And Kling, for me, is when they launched in the scene, was it one or two years ago?

It's a lot of hype around them because it was a model that wasn't Western, like US focused, a model from overseas, and it blew people's minds. Wait, other people are doing this at this level of quality? And Kling 3.0 focuses a lot on video editing and modifying and character consistency.

So the model itself is fantastic. You got all the bells and whistles, you got multi-shot, you got audio, sound effects, dialogue. But the great thing about Kling is that they released this feature called Omni. And with Omni, you can train models on it for characters or for products.

And so when you prompt for a video, you can reference a character or a product in it. And that keeps it consistent, meaning that you can reuse that character across different videos. We use that product in different ways across different videos. And this for me is the power of Kling, is that they managed to have the base of a really solid model, but then have other features around it that work individually, but also consecutively with each other.

So it gives you a lot more control. And I think if you are someone trying to create an advert for a product or UGC kind of content, Kling is just the number one place to go if you really want to build something that's based on your brand. I've been really impressed with Kling this year. Amazing.

And how about your most powerful AI video generation tool out there today? So the third one for me is interesting because we're looking at models a lot, but I'm also looking at new ways that we approach. And I think Luma Agents is a new one for me in terms of how you can generate video. And I think we're all quite familiar now with big canvases where you can connect nodes together, like Comfy UI and things where you can create a workflow for a specific video that you're working on.

Luma Agents, it's an agent that you work with directly that helps you to generate the video. You come up with the idea, you provide a brief, a spec ad, as long or as short as you want, and you work with a creative agent which populates that canvas for you. It can use Luma models, it can use third-party models. This has become one of my favorite tools to use right now because it's so hands-free.

I can just focus on thinking of ideas, tell it to the Luma agent, and it just does it for me while I'm just gargling out different concepts and ideas in my head. It just keeps populating and populating. And then eventually I say, okay, I like this, this, and this. This is the focus we're going to go on.

So that for me has become like, I'd probably say my third favorite because it focuses on creating things based on all the thoughts that you have at the time without you having to physically go in and do it. And it's been amazing for iterations and amazing for drafting and coming up with different ideas. Amazing. And that it's an Asian company as well?

Luma is based in the US. They're based in a San Francisco company. Yeah, they're really good people that are there. I've been using Luma since they first launched with their Dream Mission product a couple of years ago.

And they've really come a long way. And there's good people that work there. Everyone of them has been fantastic. So what are the top use cases for AI video that you're finding when you talk with businesses?

What do people want to use it for? It's really mixed, but a lot of them are curious about it because they want to know if they can generate actual workable content. Can they actually generate something and put it online? It's still a hybrid process.

And I think despite how good the models are and how great they look, I find businesses are using it more as a starting point, a draft and iteration point, which I really support. because if they're going to go work with a creative agency or creative people to film or create something, I'd rather they go in prepared with a bunch of actual visual concepts rather than just a Word document full of, we need this, this, and this, and no images and no visuals. What AI, R&Videos allow them to do is come up with the ideas ahead of time within their own company before going out there and saying, okay, we need this amount of money to make this, or this is the visual that I want.

it gives them a lot more control. I speak to a lot of smaller companies and startups. They don't have the big budgets to bring in design teams in-house and things like this. So AI video tools are a way for them to dial in their brand a little bit before they eventually go out and bring in professionals.

And some as well want to use it fully. I want to use AI video for all my ads. And that's also something that I think we're getting closer to being able to do. But I still always recommend the hybrid approach.

You still need an expert guiding the wheel a little bit. In your consulting work, do you find that people want you to do the project for them or they want you to train their team to understand the tools and for their team to be able to use this stuff? I specialize in the educational side. I want to show people how to do it for themselves because I think it's more cost-effective for them.

I think it also introduces people to how accessible these tools are and how they can jump in and use them and they don't need to pay someone thousands and thousands of dollars every time they want to create something. So they always want to see results. So I create a lot of videos based on what they're looking for or show them the limit of what I think it can do. And then after that, I say, well, why don't we sit down for another hour or two, two more sessions?

And I take you through the process of how I did this and the tools and the prompts and the process I took. So next time, you know and have that context So we have a work again you know how I work but also you can do it yourself as well And I happy to keep consulting on the side if needed but the educational bit is where I have the most passion for So I always try to push for that. I think it's a bit more of a sustainable approach for a business if they can take the time to learn rather than just expecting someone to always be there to make a video.

So it's a bit of both, but I selfishly push the education angle a lot more. Yeah, your content's great. As a content creator, what bars are you holding yourself to as far as how much you're producing? I see everything I post as my portfolio.

And this is something I try to explain to people because I do post a lot and I spend a lot of time doing content. And 90% of what I post is not paid ads or some of them are sponsored stuff, but most isn't. And the reason for that is I have a passion for creating tutorials of how to do things and reducing the gatekeeping of accessing these tools. So everything I put online is a portfolio.

So if any business or any person wants to know what I do and see examples, they can practically see hundreds to thousands of posts of tutorials and content for the last three years of AI, art, image, and video tools and processes. And that's the standard I hold myself to. I need to stay active with the space. And I've found that speaking to people and speaking at events and speaking to businesses, the hardest thing they find is keeping up.

Yeah, definitely. And people can find you everywhere on Jared Liu, your name, right? L-E-W. you yeah yeah so i shifted everything to my name jared lou uh a couple years ago it was one of the smartest decisions i ever made now everything can be found under that and i branched out this year to a newsletter to some other different platforms like threads just to give myself more opportunity to talk about content more and in a different way it's very different to record a video and to write it.

And so I thought this is the both. I do want to ask about the sentiment towards AI videos and how that's shifted over the last year since we last talked. Has it solidified? It's definitely changed a lot.

And I think if you look at professional studios, like look at Hollywood, there's more talk about AI than ever. But it's moved from, I think, a very anti-AI stance to a more, hey, we can use this as part of a hybrid process. It can be used in a certain way to support what we're doing, which still involves a human in the loop. And I think that level of maturity of the discussion has been a huge theme this year.

More studios and more directors are talking about it in a more productive way rather than just we hate it. I think that nuance is really important. And I think that is reflecting how the general consumer is looking at AI as well. I think there's always going to be skeptics.

There's always going to be people that are anti-AI or they hate AI or they hate the people that make the tools or they hate creators that talk about it. I think that's normal across all disciplines, especially when a new form of technology comes into place. It's normal to be a bit confused and a bit scared and a bit skeptical and thinking this is going to take over this job or remove this person's whole livelihood. I think it's normal.

But I want to be really clear that I do not agree that this stuff should replace anyone. I think any company saying that we have AI is going to replace my entire creative team. I think that's just ridiculous. And it's not the way to talk about this technology.

It's just, for me, fear-mongering. And I think the more and more I use these tools, the more I realize I need to be part of that process more than anything. and I think experts are starting to see that as well. It's not a replacement.

It's a tool that can be used to do more and expand your artistic vision. You've been in this for a long time. How good is AI video going to get? Every prediction that I've ever had on this stuff has been completely destroyed because we have 1080p video that I can do with dialogue and sound effects and music and perfect cuts.

I, there's no way I would ever been able to predict that, but after seeing how fast this has become, I think we're going to enter a time where hopefully this stuff gets more accessible and cheaper. I think we're already seeing that now, if you use any of the meta apps like Facebook or Instagram, we can now generate images and video as you're chatting to your friends and family now. So it's becoming a lot more consumer friendly, which I think is really cool. But I also think we're going to reach a point where the models are strong enough.

And I think we're seeing that already with the CDANs model. They're strong enough, the quality is there, the character consistency is there, it looks great. And the question will come down to how do we use them and what's the best way to use them? And I think that will be the next competition across all these AI video models and platforms is what can we do to make our model as accessible and easy to use as possible?

whilst also letting experts come in and make those fine-tuned adjustments that they need. I think that's the next fight. Because when you have a great model in place, you have to start focusing on how people use it. What's your view on these video generations getting so realistic that you can't tell anymore whether it's real or AI-generated?

What's your view on deepfakes? Deepfakes, terrible. Anything to do with AI and misinformation and disinformation is wrong. And I've just spent the last nine years working in crisis communications and social media, brand reputation, stuff like this.

And deepfakes misinformation is very hard to combat, very hard to identify, and also can really ruin people's lives. And so we need to have more discussions on the ethical side of things. I think labels are sure one thing we can do and should do. And companies like Adobe are doing work in this space with their Content Authenticity Initiative where they working with different platforms So if you do post something an image you can trace it back to the source where and how it was created But I think there's an opportunity now for us to have more transparency across all forms of creativity, including writing.

I started questioning, we should start questioning everything that we see online. And I think we need to get to the point now where if I read something, news, a post on X, something on Instagram, I think we need to try and give more information as to where this has come from, why it was created and the process for it. I think it will protect artists more, especially the ones that are not using AI and are worried that their art is being stolen and trained for models. I think we need to encourage people to share more about their processes.

So it really shows that this is a unique individual product. and I wish it wasn't this way because we have a choice whether we steal something or we deep fake something but I can't stop the trend I can't stop the wave of people doing this but there was an opportunity to do the right thing here and I don't think many people have done that unfortunately it just leaves a big stain on this technology as a whole and honestly it's a big distraction from the potential of it so I think labels in the short term is good but I think we need to encourage more transparency around what we're seeing online and how it was created and why it's there.

I think platforms will have rules that we have to follow, obviously. But I think as creators and anyone who does stuff in this space, I think we also have the option to do more ourselves, to make it aware that this is AI or how it was created. A lot of the videos and tutorials I make, the reason they're a little bit sometimes low quality because I'm literally screen recording, is because I really want to show the physical process of how something looks when it is created. And that's why I've swung away from getting a graphic designer in to design something that doesn't reflect really how it looks.

And that's one thing I'm trying to do to be as transparent as possible in my content. And then sure, I can label on X made with AI or go to YouTube and say, all right, I have to put this label by AI. Nothing's stopping me from putting more information in that description box and saying how this was created, where the inspiration came from. I want to give you a chance to talk about personal AIs and you're running your own LLM on your own device or you're figuring out how to do that.

And just give us layman's, what are the capabilities today and what are you doing? OpenClaw was a big thing. What that unlocked for me is obviously having an AI that works with our files and works with our information, connects to our emails and stuff. But for me, the thing that really stood out from that whole discussion was being able to run all this stuff on our own devices, which means you can run it without having any connection to the web, which essentially hopefully means that the data is safer and we keep it within our own walls rather than siphoning it out to a company like OpenAI, like Claw, like Google.

And I think in the future, the only way to keep safe is to be able to run this stuff locally. So on my phone or on my laptop. And over the last couple of months, I've been digging as deep as I can with the time I have to what do I need to run a model locally? And you don't need that much to run a model.

It's just more if you have a weaker device, you'll have a weaker model. But you can still see that potential of, OK, cool. I'm on an airplane. I can still chat to my LLM because it doesn't need to be connected online.

And so I've been doing a lot of testing with OpenClaw, but also new tools that have emerged, such as Clawed Cowork and Clawed Code, to really test in a safer way. How much do I trust giving an AI my information? But also, how good is it really? Can it really write a newsletter for me?

Can it really read my files and make decisions on if something should be deleted from my computer or kept? Do I trust this tool to read my calendar and emails accurately? So that's the current process of where I am now is trying to just see how much I trusted for making bigger investments to get a bigger computer that could run this stuff locally. So what are your first reactions to it?

I mean, does it seem like the hype is worth it? Is it making these decisions well or is it still a little clunky? It's a bit clunky, but the potential is there. Sure, it can't write me the perfect newsletter, but it can give me a pretty good draft of a newsletter.

We're already at this point now. My hope is that the next wave of personal AI, whether it's this year or next year, is stronger to be able to run on our smartphones, something that we can just hold with us. And I think that's going to be the number one way to protect our information and our data into the future. And it gives us control rather than always having to rely on a cloud service from a different provider.

I worry about people losing their data. I worry about being hacked. I'm worried about all these range of things that are really out of our control. But if everything is kept locally on my devices, it feels a lot safer.

And I really hope this is the path we go down. Jared, this has been a fascinating conversation. I'm really excited to get this out there to people. Is there any parting message you'd like to say to our audience?

Yeah. First of all, thank you for the opportunity. Always a pleasure to speak to you, Chris. And I think the main message I want to give to everyone is just this is a time where I really recommend you stay curious and stay active with all the changes and updates that are happening in the AI space, not just with art and video, but in general.

I think this is a really big year where we're going to see some major launches in the space, but also in the personal AI space as well. So I encourage curiosity this year as much as possible. and don't be afraid to dive in and try new things. This is a time.

It's a beautiful time to go and do it. Amazing. Jared Lube, thank you so much. Thank you.