42: Using AI for Generating Video Content with Michael Powers

This podcast episode is sponsored by chiefaiofficer.com offering training and certification through the International association of Chief AI Officers. Interested in a new career or leveling up your value in the marketplace? Chiefaiofficer.com can help. Welcome to this week's episode of Using AI at Work. Our guest today is Michael Powers, founder of Twisty AI. Michael's been a pioneer in digital video and now he's focused on AI generated video being as accessible as any other modality of generative AI. It's a fascinating discussion with one of the top minds in the space. So let's get started. Welcome to Using AI at Work. I'm your host, Chris Daigle. Each week we'll be learning how today's business owners, entrepreneurs and ambitious professionals are getting more done with smart use of tomorrow's tech. Let's get started. Okay, everybody, welcome to this episode. We've got Michael Powers, founder of Twisty AI, as our guest today. And how we got to know each other was through a mutual friend here in Austin, Texas named Nick, who said, you guys ought to get connected because one of the things we do with Chief AI Officer is we host a. A monthly get together for AI founders and AI people that are doing stuff in the space. Not just casual folks, but people who are actually building in the space. And Michael had a chance to join us last night and I just found out Powers had a good time. So we're going to have a conversation today just about how he's using AI at work in his business. So, Powers, this is not your first effort. I mean, first, I guess. Tell us about Twisty.

Yeah, Twisty is putting together or sort of unlocking the power of video AI, which is just coming into its own now, mixing that with language models which we know are very powerful, and putting it together in a way where small businesses can sort of have this intelligent agency at their disposal to help them do social media strategy, help them produce daily videos that are on message for their business and have that content ready to go without the need for a lot of editing to post to all their social channels. So we're just basically trying to just unlock all the creative power, I think that is in language models and increasingly in the video and the music models as well.

And the user doesn't have to be the creative one.

Yeah, they're the idea person would they normally would tell an agency like, I'm thinking about this messages around my wellness brand or my juice brand or my, my healthy soda pop. And then the agency has to translate that into Content programs for them and do the back and forth and the crafting and everything. So we're trying to shorten that life cycle and let the ideas from the owner of the business and if they have a social media person just be able to put those ideas in and then get videos that steadily come to them that they can shape and direct and say, yeah, that story is good, that that factual video is really good, or I need to change it, I need to adjust it. So it's sort of like a agency partner, except you don't have the high cost and you don't have the long turnaround times as well.

You know, when people hear AI for video, you know, they ask. I'm sure they ask you, because the first thing I thought was this, is it like avatars? Is it, you know, is it. Are you making movies? So help the listener kind of understand what you're defining as AI for video and what the output is from Twisty.

Yeah. People who are familiar with a lot of the agent talking, talking, I guess, talking heads videos that you can use, they're powerful things like hey Gen are really amazing packages. Our focus is on storytelling. So it's basically creating content that's a. That's telling a story. It might be an inspirational story, it might be an informative tips and trick and tricks or advice story, but they're 15 seconds to 30 seconds long. Multiple scenes, usually around characters that are saying, you know, here's a story about Mike who woke up in the morning and did some, you know, picked up the fresh juice, juice brand, drank it, and then had energy to do things even better that day for himself. That would be an inspirational story. And we do all the work of doing the script writing and the individual scenes and keeping it consistent and doing the captioning and the music background track so they all match to the theme, which was usually what a video editor would do for you.

Yeah. So when you say we do the work, actually you've built the technology. Yeah. With AI actually doing all the work. So when I got home last night, I set up an account with Twisty and I actually created our first video. So it's very cool. And we do a lot with social. And I've actually got my chief content guy like, dude, check this out. How can we plug this in? Because we do a lot of. We do a podcast, authentic conversation. You and I have a conversation. And from there we look to use all the different generative AI tools with different modalities. Voice, a video, blah, blah, blah. Not video yet. Except for clips Right. Like the opus and things like that. And what I want people to know is that this is not that this isn't some clip that you took from something else. This is, this is unique generated AI video that's specific to my brand. My. The type of message, like you said, is it informative, is it inspirational and that sort of thing. So it's not something that I'm seeing out there. We look at all the tools, but I'm not seeing anything like this. And is this a new category that you guys are creating here?

I believe it is. You can look at some of the demos like when OpenAI showcased Sora early and kind of put their place in the map and said video AI is coming. It's not live yet on their platform, but other services like Runway and Pika and a few other ones have some pretty good models. So a lot of the showcase videos they show are short stories. They're showing stories of characters doing things. So applying that to social media marketing and making engaging content that just grows your audience and just gives your, you know, gets your messages out there and creates searchable content. I think it's a new application for it and I think it's really interesting to take language models and lean on them for truly creative output. So for instance, we have a beard oil company and we of the themes they have is historical figures who discover beard oil and it makes their beard attractive and how does it change their place in history? It's a very creative, kind of fun thing. But the language model is doing all the heavy lifting behind that. It's creating three unique story suggestions every day with different historical figures based on its wealth of training. And then it's creating unique stories for each of them. So when the, when the owner wakes up in the morning, the first thing they do, they check their account. They've got three 15 second videos that are telling stories about Abraham Lincoln or Socrates or Woodrow Wilson or whoever it is. And they can decide whether the story is engaging or if it's just too far out. But the language model is doing the creative work.

So it's waiting on the, the social person, the social media person in the. Like they show up the next day and there's their content selections for the day.

Exactly. And they might run, say they run four different themes. They might, they'll wake up to 12 videos every single morning.

Very cool.

And. And we hook up to their social channel. So if they hook up their TikTok or Instagram or Facebook or YouTube shorts currently, we can automatically post it. So they just Say schedule. I like that one. Schedule it. I like this one. Let me edit the caption a little bit to bring my brand into it and then say schedule. Or they can say, yeah, that one's not quite right. Let me just hide that. So they just kind of like they go through and they choose for the day three or four pieces of content that match for them and then they're done for the day. It will, you know. 11:11pm 1:00pm 5:00pm Post infinite.

You know, one of the things I like about this is that as I mentioned, we use, we've, we've tested a lot of tools to, to repurpose content from our podcast and the training calls that we do and things like that, and most people out there who are, I don't know, leading edge when it comes to using these tools to repurpose their content, we can very quickly distinguish, like, oh, they're using this tool, they're using that tool.

Right.

But one of the things that I am not seeing is the type of output that I got from Twisty being used in social. So it is an opportunity to stand out. Obviously it's, it's AI generated and that's like, you know, sometimes you want it to be indiscenable from human output, but if it's going to be AI generated, you want to kind of lean into it. Right. And like this is what's possible.

Yeah.

And I love the fact that I'm not seeing this out there. So when we start using this, we will have a, I don't know, a point of distinction in the marketplace when it comes to all the other AI thought leaders or whatever that are dropping content out there.

Yeah. I think as part of our, as a partner service, one of the things that's on us to do is to keep your options fresh and on trend. So right now we have narrative stories, we have informative stories. So for social, you want content that's searchable after the fact because search on TikTok is a very heavily used thing. So that beard oil company has one about daily tips and tricks on keeping your beard trimmed appropriately. So that's very searchable content. And then the stories are more viral, inspirational, engagement oriented. So if you run multiple themes, you can just feed those different purposes. We're working on a product animation path as well. We, we have it working on the site somewhat. You can give it a image, a photo of your product and just say what you want to do in the background, like make a rainstorm in the background and it will keep your product tight, sharp, and then let the engine create the background for you. So that way you can animate a product. You can post it to your social. You could also take that same video and use it in your catalog if you're selling it on social. So. So there's many uses. We're trying to create all the workflows that a business owner with products or services to sell can use. You know, you want to create a Facebook ad, we can create that for you. You want to create a catalog entry, we can do that. And then we have our daily themes. And that's one of the wonders of, like, really unlocking the power of AI video, is that we get to give you the most important sort of workflows that you can use every day.

So one of the things that this is not your first attempt at video in a digital environment. Can you kind of share how you got on this path to twisty?

Yeah, yeah, 100%. So definitely a startup, whatever you call it. A very seasoned startup person. I lived in Silicon Valley for quite a while, and this is my seventh startup, actually. Nice. Very fortunate. On the third one, first two were learning experiences. On the third one, I got connected to Max Levchin and the PayPal group, who had an incubator that he was personally building up companies. So in that incubator, there were two people starting what would become Yelp. There was myself and Max, who started a company called Slide. And there was another company doing, like white hat SEO stuff. So he had three companies and he put a million into each of them and was trying to get them off the ground. Very quickly, the SEO one went belly up, and those people just kind of disappeared one day. And we kept working on Slide that it funded. A couple months later, the people from that company that did not make it called me up and said, hey, we're making a video site. We need a product manager. We loved talking to you in the office. What do you think about jumping from Slide over to here? Well, that company was YouTube, and it was. It was very early, and I love video. And so I was like, yes, I'm going to do it. So I jumped over to YouTube, became the second product manager. My pitch to them was, you know, everybody should be a channel. That was my pitch.

Like, interesting.

At the time, it was all viral videos. So my pitch was like, hey, I'm signing this series of content, comedic content or video blog content. So I said, you know, let's make a channel and let's make it so you can subscribe to that channel. So the first product I Worked on a month after joining was YouTube channels and YouTube subscriptions.

So where did you go from YouTube?

I stayed at Google after they bought it very quickly and moved to the revenue side. Learned a lot about sales, a lot about selling the brands and businesses. Built a contest platform that was really good at taking the wild creativity of YouTube and making brands feel safe our around it. Yeah, yeah. So we built that up, built up revenue very quickly over there, learned a lot. Then I went off and I started another company which was sort of, sort of like Discord for mobile. Google backed it. We had a bunch of games that mobile games that adopted it Turns out Discord took the right path. I, I took the wrong one. They took the right one and went PC oriented. They didn't have to worry about anybody adopting anything. I had to sell one by one games and get them to integrate. Yeah, yeah, Discord's a great product and that was definitely where I was chasing that as well. Then I went in house for CBS Interactive and I ran their entertainment and gaming division to kind of revitalize that whole area, especially on social. And what I learned there was, you know, we got, we got to 2 billion video views a month on Facebook and we did it in about a year and a half. We were number one and the number four, channel gaming and movie content. So it was expensive. You know, it cost us three, three and a half million a year just to produce the content and feed the algorithms. But we get, we did get to scale. We were able to sell against that. But taking that lesson, I've been on both sides. I help with the YouTube algorithms. I learned the Facebook Instagram algorithms really well. What I learned is it's, it's very expensive, but very rewarding if you can create steady content on those platforms. What they want is engagement three, four times a day. If you can do engaging content that often. And you know, brands like Mr. Beast know this very well. They play this game all the time, just a drumbeat. You just want to create engaging content. The algorithm will introduce you to people who have never seen you before. And that's how you broaden your audience very rapidly. So one of the things we're doing at Twisty, one of my learnings was, hey, video is expensive, but if I can unlock the power of AI to be this creative powerhouse for businesses, then I can give them three great pieces of content every day and they can scale their business and sell more product and make more of an impact. So that's really where our mission came from.

Are you guys doing that yourselves? With Twisty, that frequent tempo of creation and posting.

Yeah, we've started to. We sort of released this automation towards the end of August and we just got approved for some of the social platforms but we've been scaling on all four of those channels currently and making good progress. So yeah, I always test out the themes ourselves and yeah, I'm going to be doing really well on YouTube. Shorts and TikTok actually are the two that we're growing the most quickly with.

The AI generated videos.

Yes, all AI generated videos.

Y do, do they have the ones that are getting the traction for you? Do they have narration or is it just the music and image?

It's just the music and image and I'm finding on TikTok the trend has been away from voiceover narration. There's still some like historical pieces still tend to carry that. But in general music has kind of been dominating and just animated text on top.

Yeah, yeah.

So that's the format we've been leaning into. You know the trends may shift and in that case we bring in a voice API to kind of do redetects that the story is nice. But yeah, we have a lot of work to do to like give you different font animation options. There are sort of like three or four major ones that are really readable and good and visually appealing. We have one option currently so we're working on that right now to give you about two or three more options. And the background music, we're using something out of Facebook from a year ago. It does, it does okay on background music but the current generation, you know, Suno and Audio are amazing and they have APIs so we're looking at them. It's like okay, let's, let's lean into that and do our next gen background. But we've done thousands of video posts and had 00 be flagged for anything.

Okay, cool man. I'm thinking about this and like as I mean a startup, there's always a huge checklist of things that you have to do. But man, I'm seeing now like you could get to the point to where you could plug in something like a text to voice tool and you could plug in a Suno and you could, based on whatever the output was from the video, you could have that custom song or the spokesperson's voice over the custom like I mean it just gets.

Yeah, exactly.

The only limitation is how clever you are I guess in doing this stuff.

Yeah, it's all workflow and we try to like there's that balance of simplifying and Automating. Yeah, and just. Just going between. And you can do a little polish at the end. But we're trying not to be just a pure video editor. There are some companies chasing that, but Adobe's doing that and Canva will as well. But there, the focus is on let me take every piece and let me polish the video and let me do cuts and all that. And we're trying to just as much as possible, rely on the AI layers to do that for you and get it mostly right the first time.

Man, I got a strange question. So, you know, when people generate images, images aren't as. As crazy as they used to be. They're. They're actually a lot better. But with video, AI generated video, sometimes you can get some weird stuff, right? Does. Would it be a bad thing to necessarily post those? I mean, because here's what we want, when at least what we want, we want people as they're going the, you know, scroll, scroll, scroll. We want them to be like, what is that? Something that's novel, something that stands out.

Right? Yeah.

And those weird videos that, that AI can only generate. Right. They are certainly something that, to me, I'm like, what is that? Right. So it gets. It gets the reader's attention to stop. It's not necessarily that they're stopping because, oh, my gosh, this video is breathtaking. And it's because this video is weird. What are your thoughts on actually using. I know that the models, you can't always predict what comes out, but if there is that kind of odd video, is that okay to use in my social posts?

Yeah, completely. And you can push that. You can. It's sort of like you go back, you know, go back six months and nine out of ten. Nine out of ten videos would be in that category of like, oh, great, until their head turned into a lemon watermelon. Ye. And then you're like, oh, maybe I'll keep it. That's great. Yeah. Now you mostly kind of get what you ask for. And so then it's a matter of sort of bending it and saying, we'll pick a style. You can pick 3D animation or anime or graphic illustration, and then you can mix the topics. You can basically say, you know, the characters can be animals. It can be, you know, a raccoon doing different things, or you can do a mermaid thing, or you can, like, do underwater stuff. You can basically do weird things that you can't shoot in the real world. You can just mix, like a panda bear swimming with scuba gear. You can do that. And it will be realistic. And so then you can have a lot of fun with that. And then there's always the unexpected. Like I did one the other day and it was more of a fantasy one. And at the end, the guy, you know, picks up an ax and goes running through the woods. Well, the axe was like the size of a house. It was just this giant ax. And it's. Because I didn't specify any details, I just said he picks up an ax. And so, you know, everything matters with this. Yeah, yeah. And it's been the current generation. What's nice is, oh, here's, here's an interesting thing for people who are playing, who are used to playing with ChatGPT or, you know, different chat models. You're used to having context, right? And the context has been getting longer and longer. I think now it's like a hundred thousand or a million tokens, I think. So it remembers what you've been talking about as you go back and forth. And you can refine, you can say, yes, you know, rewrite that, but soften the tone or rewrite that and make it more marketing copy. Video and images have not had that, so they've been limited to 75 tokens, which is about 60 words. So when you think about that, you're trying to describe like a poster or something. It's just 60 words, like every word matters and you can't fine tune it very well. That's been a real limiter. Now the current generation expands, that about doubles it to about 122 words, which we have an enhancement layer. So when, when the story gets split out or you put in a story, you say, I want a panda bear who swims underwater and discovers a shipwreck. You and I would probably do that and not give a lot of detail. And you hit enhance and then it makes it 122words. It describes what the panda looks like, what the background, what the water, what the. What kind of ship is it, what lighting, what camera move, and then that goes into the video engine. You get a much bigger, better video at the end. So having more context is helping, and having better trained video models is helping a lot as well.

So that's not a limitation on your side, that's a limitation on, on the models that you're using. We'll be back after a quick break.

Do you want to start your own unsexy business offering a local service or a trade to your local community? Come check out the Owner Ops podcast. We bring on people like Sam, who has built a $22 million roofing company in five years. I've also brought on people like Brock, who generated $1.3 million of revenue in his first seven months installing septic systems. If you are looking to build your own local service or trade business, come check out the Owner Ops podcast. That's ownerops.com O W N R ops.com.

Yeah, the models, the new generation, which are based on the same stack that Sora showed, have the capability to add more tokens. So as you train it with that number of tokens. So right now most of the models are coming out with 122, but I would expect that to keep expanding and then you really have to rely on the AI to do the scene because writing a 250 word or longer scene is a little daunting. But we can take the idea and just enhance it with the right elements to tell the story.

You've touched on something and that is like routinely we test a lot of tools in our community and people come back and say, oh, I tried it. It wasn't very good. And that's not my experience because they tried it a month ago and I tried it today and it's different. Right? Like the tool has gotten better. What has that, like, what is the, the cycle of improvement or enhancement looked like for what you guys are doing at Twisty? Like what was it like six months ago?

Oh yeah, the amount of change. Okay, so we launched our first sandbox, I would say a year ago, right, Last summer. At that point, the models look like really bad stop motion animation. Just really like six frames a second. Jittery, definitely not video. Um, but we knew it was going to improve quickly. I would say coming into the, into the 2024 February timeframe, there was a jump on the motion training. Still using frames, but more motion, much better, more usable content. Just as of August this year, we have our hands on the next generation model where we're getting much clearer output from it. More controllable, less hallucination, much tighter video. Now our premise. So I would, I would say just for your audience, like 2025 to me is going to be the year of video AI. It's going to, it's going to be very front and center, very high quality. The wow factor is going to be as, as it. As. As potent as ChatGPT when it made its big splash. So I see that happening. Like it's all the models that are coming out, each one wows better than the last one. So going into next year is going to be the big year. Now our Premise is that the open model, we have an open model. That's what's running twisty right now. It's an open source model that we fine tune with some more training data. But the core is pretty good. That that model is going to rival what you saw from Sora by January.

Wow.

And that us optimizing it and putting on our servers, we can create a 5 second video clip probably for a penny and a half. Wow. So a story which is three clips will be $0.03, $0.04. Our premise is that the video is going to be very creative, look like it's shot by a real person, very usable and cost you nothing to make it. And so we keep thinking like, okay, that's our given. So how do we unlock that power for the business owner and the top influencers to use that creatively when the whole game changes, it doesn't cost you thousands of dollars per video, it costs you 3 cents per video. So we're kind of, we trying to live in that future as much as possible. That's why we're doing daily themes. That's why we're giving you 15 videos every single day. We're trying to live in that future where you're not handcrafting nitty gritty little pieces. You're shaping and directing. You're putting ideas and shaping and directing and selecting. You're really good at selecting stuff. So Nvidia's cheap then that should be your universe. Your universe should be ideas, themes on trend, strategy, instantaneous. It should be instant, not a week to get there. And that's your new reality. That's what it means to do video content marketing.

So it will be a skill that has nothing to do with the technology and it has to do more with the instruction or the input that I give the, the model.

Exactly.

How does somebody learn that? How does somebody get good at that?

Yeah, that. I'm watching that with our clients already. This whole idea that I can do a theme that I can say like we have one client who's doing a lot of a marketplace for kids creative after school activities. So he started with one theme that was kids just kids put down their iPhone and go out into nature and discover something creative. The very short prompt. And he was getting okay stories out of it. But then what we did is we took, he had a brief on like what his business was all about. And he had a paragraph, it was probably like probably three paragraphs, four paragraphs long. That was like all the reasoning on why you should put down the device, why you should go Outside how it helps you with art and music and. And speech and storytelling. Very long. We just took it. We just cut and pasted it and put it into the prompt and added a couple things like no teenagers, no toddlers, no school settings. You just wanted it to be outdoors. The videos coming out of that, on the mark, like every time. Nice, pretty happy. And that was an eye opener for me. I'm like, oh, this is the full power of GPT, right? We can take a very elaborate description of a theme and then say, given that, write compelling stories and then we know how to translate dynamics.

All that within Twisty. Right now, like, as a user, I could do that stuff now.

Yes. Yep, Absolutely.

If you're enjoying this episode and want to learn more about how to start using AI at work, we've made it easy for you. For just $1, you can have full access to the Chief AI Officer community, which will give you additional training, custom software, daily Training, calls on AI tools, using AI automations, getting more from your ChatGPT sessions, and the business of being an AI consultant. Simply go to chiefaiofficer.com insiders to accelerate your AI journey. Now, back to the episode. Interesting. Well, I'm excited. We've got a very young and talented social media guy who's loving using all these tools. And anytime I bring up something new, he's all about it. So I would expect that we're going to be able to maybe show you some interesting things.

I feel like we've only scratched the surface, like just unlocking that power and seeing how the clients are using it. I've been able to lean in with many of them and help them shape their first theme. So I'm learning as well. And it is different. It's different than writing an individual story for one video. It's like, no, I want a series. I want a series about wellness tips for men, because I have a men's brand and I want the wellness tips to be about how to get more energy and how to kind of feel stronger throughout the day. And that's a different way of thinking. Right.

You know, one of the things that seems to be, and this has always been the case with technology in my experience, is that the marketers tend to kind of figure out how to use it first. Right. Because they're always looking for an edge and that sort of thing. What are some of the opportunities outside of marketing? Because social, you know, I guess some people could do it for fun, but most of your clients, I would imagine, are doing it to establish brand awareness. And product awareness and stuff like that. Do you have people doing it in ways that are surprising you, that aren't necessarily related to marketing?

Yeah, there's. There's been a movement on TikTok in particular somewhat on Instagram of doing faceless content. So whole channels where no one talks to the camera and just building up, you know, TikTok has a great creative program, so if you build up a certain following, you can monetize it. So there have been a bunch of tools that have been trying to address the faceless content world. And so we have quite a few customers who are purely. They don't have a business, they're just creators. And so one, it's on the political side, but one, you know, the whole cat lady thing. So one person set up a channel which is Cats for Harris, and just does all kinds of cat videos with little saying on the shirt. So they're taking images in from mid journey and then feeding him into our thing to animate them. And they're very popular, they're very viral. And that was fascinating to see. And I'm seeing I have a channel that I set up that was just. I just noticed that videos with animals were doing really well on TikTok and then. And in particular videos with wolves were doing really well. So I just set up a separate channel and I just feed it with three wolf stories every day. I just wanted to see how far it could go.

Sure.

Yeah. Very popular. And it, it serves no marketing purpose. It's just. It's a niche. It's a niche. So that is untapped. Like that's interesting. Art tool can be used that way.

Yeah.

And. And it's a valuable, it's probably a valuable arm of how you would use this creative storytelling platform.

Do you have an environment where your users can share kind of like what's working for them with each other?

Yes. Yeah. We still kept this sandbox that we originally made. In fact, we have an iOS app, we have an Android app that gives you access to everything within that you can hit. There's a Create button. And the Create button lets you make individual videos or individual stories or image to video. You can do all the fine grain stuff and then we have the themes. Now we have trending videos. So most people leave things public. You don't have to. You can make it private. Yeah. But we, we have public profiles for anybody who has an account. So if you're a customer in a brand, you can just have Twisty be an additional social media account for you and publish what you want there. And then we have upvotes and following and all the typical social media things and we play back the videos TikTok style in the scrolling. So and we do, we do challenges. Every three days we do a thematic challenge. And that just keeps the creative juices going for people who are trying to create interesting content. And so yeah, that's all there. And I think, I think that's an interesting test bed. It may grow up to be an actual social platform with some, with some value to companies, but right now it's a good creative space to learn from people. Yeah. And everything's shareable. You can share the videos, you can post them, do all that.

One of the things that fascinates me is the lack of learning curve required for most of these generative AI tools. Right. I don't have a video background. I told you I tried to go to the studio. The. My friend who runs. Runs the studio wasn't there. I had no idea how to. Now I'm back in my home office because as a result, same thing with like, I can prompt and I can go to a tool like we use a lot internally. We use Leonardo for our still images and stuff like that. I can get good output from that. But video, to me seems like I just don't have a lot of experience with that. What is the learning curve for somebody who wants to, to do that? They want to have those, those videos generated on a daily basis so that they can introduce them into their social. What's the turnaround look like for somebody to sign up and then be able to go, oh, wow, this is good.

I think with themes, it's pretty quick. I've onboarded, I'm continually onboarding companies and so I usually sit with them and hear what their messaging is around their business or service, their product or service. And then they usually have ideas. I'm like, well, what kind of themes would you want to put out there? What kind of videos would you want to put out? And they'll, their team will usually put out a bunch of ideas or they have a written paper on, like, this is a kind of. This is our philosophy. And so from that I help them set up their first four themes and then we run like with your first themes, you can just press a button, you get the first videos right away. So I run that, I have them look it over and tell me if any of them match.

Right there on the call.

Yeah, right on the call. Yep. And any of these match what you want. And then if something's getting close, we're like, well, let's lean into that theme and let's shape it a bit more, let's make it closer to what you want, but let's get rid of a theme. It didn't match anything. And so usually within the first week with that back and forth a couple times, they'll start getting videos that they're, I can always tell, they're always like, okay, now how do I hook up my social media account? Because they're ready to post it. Yeah. And then usually by week two, they're leaning in and adding a new theme or shaping a theme. They're kind of off and running and I love seeing that because none of them are video editors. Usually by and large they're, they're just getting good output. And now for the video editors, the people who like to craft a story in more detail that's available, you can just jump into the story creator, detail out your story prompt and then that will turn into a 15 or 30 second video. You're still not video editing, you're still just describing. But if you're a good storyteller and you know you like, do like, it's equivalent to using an image prompt.

Yeah.

You've got to put in things like camera panning, camera overhead, what kind of lighting you want. You gotta, you gotta think about that stuff. And then hit our enhance button. It'll kind of polish it for you. So that's if someone's comfortable image editing, then I'm like, okay, go ahead, try make a, make a story. You might actually get good results.

Yeah, I'm excited for us to start using this. So as I mentioned, you know, one of the things we like to talk about is how you as a business owner are using generative AI not exclusively like your tools, but like what does, what are some of the common tools that you're using just to run your business more efficiently and effectively.

Yeah, I definitely use ChatGPT every day to.

Are you using the other LLMs? Are you using Claude and Gemini?

Not currently. In fact, our whole system uses the API for ChatGPT to do all the layers and we definitely work closely with Amazon and with Google. So there's always nudges to go use something that's hosted there. Yeah. But haven't found a real competitive advantage to, to shifting to one of them. So we spend a lot of energy. Like we host our own video generator and we host our own music generator. So the media side we want to have our hands on because there isn't one anything available that's really the right scale. Yeah. And then Personally I'm using ChatGPT a lot just to everything from search to polishing outbound emails to customers, investors. Our terms of service were written first with ChatGPT and I gave it a reference of a couple other sites like YouTube to reference off of like really valuable on that end. And then I do spend a lot of time exploring, especially media generators. So I spent a lot of time in Suno the last couple months to see what the capabilities are there. I spent a lot of time in Runway and all the video editors, I just want to see like what they're good at, what kind of content they make.

Well, what are you discovering like with Suno? And for those listening, if you haven't used Sunosuno AI is the music generates songs.

Yeah, yeah. Suno and audio, I think adio. They're both really, really good. First of all, the wow factor of I can just say give me a reggae song and have it be this theme and just give it a title and I get a full song back just as like, wow. Yeah. Unbelievable. And you can get more detail with actual text and do more writing on your own. So that's a big shift. And then I'm mostly interested in background music. So then I'm like, okay, what can I do in background music and what keywords can I give it to get something really solid? And it does, it does really well with that. You have to kind of give it, you definitely have to give it a style direction. So you have to have, I think on our end when we use it, we'll have to have a list of like 120 very interesting background styles and keep it in that tight band. But it's interesting to see the next generation. And then I do play with the audio, synthetic audio, to see how natural the voices are sounding, how natural the agents are looking. I have had people, creators on our platform do a story with a character they kept consistent and then they take that video after the fact and they use hey Jen or couple other services do lip syncing now. And then they, they actually have that character in the video talk.

Oh, nice.

And that's come out. That looks really good. It does all the head movements. And you know, these faces aren't just looking head on like the original stuff. They're, they're in motion. They're, they're, they're moving around the video. So that is fascinating to see. Yeah, I think I'm, I think I do a lot of that experimentation because I want to know what's coming I want to know what the cutting edge is. I also am on high alert for any workflow. Light bulbs that go off. Like, I know what businesses are reaching for. And I'm like, how do I any of this lead to a workflow that I can simplify and put in their hands? Right. Like my current one is product placement. I hear this a lot.

Tell me more about that.

Say I'm a gin brand or our beard oil friends, they. They've got very nice packaging. You know, the beard oil, they really want it integrated into their stories much as possible. We haven't figured out a full workflow for that yet where you can actually have the product move. It'd be crisp, crisp, readable font, like stay the product and move in the action. The best we can do right now is we can animate a photo shot and we can keep it. We can keep it still and change the background. So I'm on alert. Like, who anybody has unlocked this? And that's another reason I'm a believer in open source is that that innovation happens there. They break down a workflow. Have you heard of a thing called comfyui?

No.

Here's a new one, maybe for your audience too. ComfyUI is it runs on a gamer PC, gamer quality PC. It lets you take all the modules, image generators. They do pose adapters, so you can have somebody dancing and it will figure out the pose of them. And you can say, make this a panda bear. And the panda bear is doing the exact moves that that person did. You can do face transfer, which we.

Have with that, that viral video of the guy going out on stage singing to the audience, and they had the joker. Was that. You know what I'm talking about? Yeah, I was done.

I was done with that. Yes.

Yep, got it. Okay. Yes, I know what that is. It is pretty cool stuff.

So that's where all the innovation on the. On the workflow frame control is happening, because they can get their hands into the code and reveal every single frame as it's being generated and they can modify it. So we've had it on a site for a while. We had face transfer. I could take a photo of you and I could say, make you climb El Capitan or make you do a kayak down the Colorado river and have you skydiving out of a plane. And it would work. It would be your face transferred. Yeah. So that hasn't. In the new generation of video, that hasn't. Hasn't been exposed yet. But it will. It will be exposed, and then we'll have More options.

So you mentioned 2025. What do you think is going to be some of the breakthroughs, at least for the AI generated video in 2025?

I think the core model is finally. Core models are finally going to come out where, um, they're very well trained, generally sort of the equivalent of ChatGPT. Like they're good at just about everything. And then usually what happens is these models are open to fine tuning. So the general model will be good, but it won't be great at anime. So someone will take it and fine tune it for anime or 3D, Pixar style animation. Right. Those are usually specialties. But the core model is going to be very capable and very flexible. But then the innovations I think that are going to happen on top of that are multi scene cuts with really good camera control and really good quality cuts between scenes where you can get a five second clip and it has a couple of shifts in the camera, it's not just one shot. So I'm seeing some models starting to be trained on good content so it knows that. The other thing I'm seeing is syncing it with foley sounds. So the door creaks open, the car starts, the engine revs. Like I'm seeing Facebook has a new model they teased that actually does all of the foley sounds. And that's a, that's a whole new world. And then conversational piece, we'll start to see some models that mix the power of the synthetic audio with actual conversational expressive faces in scene and full gestures of characters, actors in it. Because right now that doesn't exist. So all that will happen in 2025. We're gonna see like one piece happen other.

I was visiting la, I guess whenever the strike was happening, right?

Yeah.

And I was driving through Burbank and there was everybody in front of the studios with their signs and they were. It was a lot of that had to do with fears about AI get being involved in the usage of the production of entertaining entertainment. Right?

Yeah.

Do you. How do you see that pushback? Because if it continues to get better like this, I mean it's business, right. It's not sure there's people out there that have, have spent a lifetime developing a craft, but for that expertise, a company will pay a premium for that. And if that 8020 even can be produced at like you said, 5 cents a video, what do you see that doing too? I mean, and I hate asking these questions because I don't have an answer for them, but I know that it is something that people think about, like, what happens when AI can do blank and we don't need as many button pushers or whatever.

Yeah, yeah, I think about that a lot. This is such a new area. And copyright law hasn't caught up to it either. I did see it in negotiations. I mean, if I'm. If I'm an actor, you do want to build into your contract. Like them building a synthetic you and putting you in other stories, other. Other movies and TV shows like that has to be protected because that's your. They're going to cut you out at that point if you. Especially if you're not getting. You have an abrupt. You're not getting residuals from it. Right. You're not getting paid for it. So I think there's that individual kind of.

They're now saying, like, I own any synthetic representations of me or.

Yeah, the studios would. If they could get away with it. They would kind of say, hey, yeah, we have all this footage of you. We're going to train a model and we're going to extend the director's cut of Blade Runner. Or we're going to. You know. And they have. They've already done that. Right. A young Luke Skywalker and the new.

Yeah.

Rebel Maroon. Right. Or whatever it was. They already have that capability. So it's taking it to the next level and saying, well, what if, you know, if they capture so much of your speech and your gestures and everything, and then they can actually make an entire film without you ever showing up to the set. I think that was one of the things they had to protect is like, well, just like streaming, like, there's going to be some notion of, that's my ip, I got to be rewarded for that. That's one aspect. But the other one that I'm interested in is. So this is just. I'll give an example. I worked for Xerox for a long time. Xerox Research. Xerox was very wealthy company. They set up research in Palo Alto. Right. A lot of things came out of it. They invented so many things. The desktop computing, laser printers, Ethernet, sending documents around, all invented there. Meanwhile, Xerox is making its money off of copy machines. I worked at the mothership of Xerox. They made 98% of their money off of selling paper and toner. Not the machines. The machines they leased for, like, nothing. Yes, all paper and toner. So when this little group out of Palo Alto was like, the future is paperless. The future is going to be a paperless office. The rest of the rest of Xerox Is like, oh no, that's not going to happen. We're going to squeezed it, contained it. They're like that paperless office thing cannot be allowed to be real interesting. But the reality is when that office became a reality, people actually printed more, not less. It's just an unintended. The unexpected kind of consequences.

Yeah, yeah, yeah.

So my thought with like my maybe optimistic belief is that when we can create more content and better content and be more creative, that just creates more opportunity. I think there's more consumption and more engagement and more outlets and that just creates more opportunity, I think in the end for more people, not less people. And I know it's a very optimistic view, but it's like. And not that this, it's going to stutter into that, but I think that, I think we've seen that already with online content. Right. The fact that you have all these. Most of the watch time now is, is online on apps, not on tv. And so that shift has happened already for.

You mean like major production entertainment stuff is now being consumed on the device rather than on a tv?

Yeah, in short form. And you know, people are using apps rather than just just streaming, just like big TV streaming. I think just this last year we passed the point where that's now the dominant consumption path is through the apps, not through the streaming. And you could argue that that's created a lot of opportunity for more TV shows, more series, more experimentation. Netflix is doing a ton of that. So I can see how it's sort of like, I don't know, I think there's a. There's a basis to say that it's an expansive area and that if we get more people creating and turning their imagination into interesting content, there's just going to be more opportunity.

I love the way that you're looking at this. That's awesome. That's actually when people start asking me that question, I'm going to really explore what is the optimistic answer. Because you're right, people are. I mean I'm. I've got a printer sitting right here. Paper didn't go away. And if more people can have an idea and create those with very little barriers to getting there, I would imagine there is going to be an increase.

In that been a personal mission of mine. And when I, When I joined YouTube, I said like, you know, channels and subscriptions was my first, the first thing I built there. And it was a notion that these creators should be able to make, find their audience, make a living doing that if they have the, the ability. And that's what we're seeing. I think a lot of creators, that's their business now and. But you still look at YouTube, only 2% of people who are on YouTube make content. Only 1.6% of people on TikTok make content.

It's incredibly like on every platform, isn't it?

Every platform, yeah. I was surprised that TikTok was lower. I thought with TikTok's built in editor and tools, I thought TikTok would be much higher than YouTube. I was shocked that it's less so. You know, as someone who's been really backing the creator economy and really wanting people to create and find a following, I see AI and AI video in particular as a way to make more creators. We all have imaginations, so take your imagination and now you don't have to know. Adobe Creative Tool set right now you have this partner who can kind of understand making good lighting and continuity and decent script writing. And now you've got that ability to kind of turn your very nice imaginative ideas into something that someone else will want to watch.

I love it. Powers, this has been awesome. I'd like for people to start testing out with Twisty. What would you recommend? Obviously go to the site and set up a free account but what would be the next steps for somebody once they do that? How would they get the most out of it?

Yeah, definitely go to the site and we have a free tier so you can do individual videos but you can also create a theme and start to generate series of videos. Look at what other people are creating in the trending area for inspiration. Hook up a social channel if you have one or set one up new if you want to. Just experiment with seeing how how fast you can grow it. And I especially would like feedback ideas, first impressions. So my email, how do we do that?

Oh just email you?

Yeah, just email me. So just my. My last name Powers Wisty AI write to me directly and all thoughts good, I love it. Or hey, it's not doing what I want or I have an idea of how to make it even better. It's early right? I call it. It's a beta video. AI is new so it's this collaboration like I want to make sure the workflows work for you for what your purpose is. So don't be shy in telling me what that is.

That's awesome. We have a pretty active community of AI enabled folks at the Chief AI Officer spot. I'm going to let them know. Hey, this tool's here. He wants feedback. Go break it please. That's awesome. Well, Michael, thank you so much for taking the time out. I know that this was kind of last minute. We just met last night and I was like, I want to talk about that on a podcast. So thank you so much for doing that on short notice and I'm looking forward to supporting Twisty. You know, you're local like we are now, and anything that we can do to support that, I would count on it.

Great to hear and it was really great meeting you and the set of people. It's good to know there are people really leaning into making good use of AI, really making it make a difference for these businesses.

Beautiful. All right, everybody, thank you so much. Go use the tool and then give them the feedback. Thanks for tuning in to Using AI at Work. Don't forget to subscribe for more conversations about how to use AI at work and a special thank you to our sponsor, Chief AI Officer for empowering businesses with AI education and training. Visit their website for a free AI Readiness Assessment and AI Strategy Guide to help you get started using AI at work. That's www.chiefai officer.com. thanks to our producer Evan Desolier for making this episode possible. Follow us on Twitter at the handle Using AI at Work and visit www.usingaiatwork.com for free resources to help you harness AI in your role.