In short
Podcast Episode Summary: Scaling Visual Stories: Secrets of Consistent Character AI with Diana Zdybel
Podcast Overview Podcast Title: Leveraging AI Host: Isar Meitis Episode Number: 138 Episode Focus: The evolution of storytelling through artificial intelligence, particularly in creating consistent character visuals across multiple scenes and expressions.
Episode Description In this episode, Isar Meitis interviews Diana Zdybel, co-founder of Consistent Character AI and Neolemon, focusing on how businesses can maintain character consistency across various visual formats using AI. The episode explores the challenges and solutions in creating cohesive narratives through AI-generated visuals.
Key Concepts and Discussions
Evolution of Storytelling
- Historical Context: Storytelling has transformed from oral traditions to digital formats, influencing business communication, team engagement, and marketing strategies.
- Importance of Consistency: Maintaining a uniform character appearance across visuals is crucial for effective storytelling in business.
AI's Role in Storytelling
- Impact of AI: Rapid advancements in AI allow for quick and efficient video content creation, previously a lengthy process.
- Challenge of Consistency: Despite AI's capabilities, achieving consistent character representation in visuals remains a significant challenge.
Solutions Offered by Diana Zdybel
- Consistent Character AI: A tool developed to address the issues of character consistency, enabling users to create characters that maintain their appearance across various scenes and situations.
- Practical Applications: Focused on aiding entrepreneurs, marketers, and creators in producing cohesive visual stories that align with branding.
Tools and Techniques Discussed
- Consistent Character GPT: A tool that allows users to create characters with different expressions and poses efficiently.
- AI Image Generation Tools: Mention of popular tools like Midjourney and SegMind for creating high-quality, photorealistic images and illustrations.
- Character Creation Process:
- Users can generate characters by specifying details like age, appearance, and clothing.
- Focus on simplicity in prompts to avoid complications in AI hallucinations.
Examples of Use Cases
- Educational Content: Diana shares an example of a psychologist who created animated educational material for children using AI-generated characters.
- Creative Projects: Discussed the potential for individuals and teams to utilize AI for storytelling, whether for professional or personal projects.
Bonus Tools and Techniques
- Animation Tools: Introduction of HeyGen and Canva for animating characters with AI capabilities.
- Tips for Beginners: Recommendations for navigating AI tools successfully, including the importance of testing and iterating.
Conclusion
- Future of AI in Storytelling: The importance of embracing AI tools for efficient storytelling in business contexts.
- Diana's Invitation: Encouragement for listeners to explore Consistent Character AI and leverage AI's capabilities for their storytelling needs.
Call to Action
- Connect with Diana Zdybel: Listeners are encouraged to reach out on LinkedIn for more insights and support in AI tools.
- Access Tools: Links to resources and tools mentioned during the episode are provided for further exploration.
Additional Resources
- Consistent Character AI: [Link](https://linktr.ee/neolemon)
- Ultimate AI Course for Business People: [Link](https://multiplai.ai/ai-course/)
- YouTube Full Episodes: [YouTube Channel](https://www.youtube.com/@Multiplai_AI/)
- Isar Meitis LinkedIn: [Profile](https://www.linkedin.com/in/isarmeitis/)
Feedback Listeners are encouraged to leave reviews and share their experiences with the insights gained from this episode.
---
This markdown file encapsulates the key discussions and insights from the podcast episode on leveraging AI for consistent storytelling through visual mediums.
Written by AI. May contain mistakes. Listen to the episode to check what was said.
Transcript
Automatic transcript. May contain errors.0:00Hello and welcome to Leveraging AI, the podcast that shares practical ethical ways to leverage AI, to improve efficiency, grow your business, and advance your career. This is Isar Maitis, your host, and I've had that episode for you that I'm personally really excited about. So I'm going to take you back in time before we get started, and we're going to talk a little bit about storytelling. So storytelling has been a part of human communication since there's been human communication. It's the way people connect with one another in the best way. It's the way you generate emotions in other people to have deeper connections.
0:36Now, that thing has evolved through the years and it started from face to face. And then we had books and then radio, which was just voice and then TV and then social media. Then really all the content that we are creating is storytelling. We're trying to influence people's emotions in order to drive them into some kind of action. Now, what does that have to do with business is pretty obvious, right? You want to connect with people in your organization. You want to connect with your internal team. You want to do team building, employee engagement, but it also impacts your external side of your company like sales and marketing, et cetera.
1:15So all of this has to do with telling better stories. But what does that have to do with AI? Well, AI now enables us to create images of videos in a much more effective way than we could have ever done before. We all know that video is the most impactful kind of content, other than obviously meeting people face-to-face, but to create video content used to take weeks, sometimes months. And now we can do it with AI very quickly and very efficiently. There's only one problem, and it's a very big problem, and that is consistency, or to be more specific, the lack of consistency. If you want your story, the person in your story that is supposed to tell the message that you're trying to convey to look the same in every frame, in every shot, it is very hard to do.
2:07If you try this, even if you try to write a very long prompt and now you change it and run it again, you're going to get a slightly different look of the person that is supposed to represent you and your company and your brand with the right colors, with the right pose, with the right background and so on. So what's the solution? So lucky for us, we have the co-founder of Consister Character AI with us today. Diana Zdibel has been at this problem for about a year now, every single day, working in the trenches with her partner, initially trying to hack it through different GPTs and other solutions and better prompting.
2:47and in the past few months developing an app that will be able to do this for us even easier. But today she's going to share with us how you can do this with existing tools. What are the techniques? What are the tools? What are the prompting? What are the tricks that you need to use in order to create consistent characters that will look as the same person or as the same character, if it's like an animation kind of character, between the different scenes so you can tell a better, more compelling story and hence connect better with both internal and external people for your business. I am personally really, really excited about this and I hope you are too.
3:27And so it's time to welcome Diana. Diana, welcome to Leveraging AI.
3:35In the next few years, AI technology will change our world dramatically. Whether you are a business executive trying to catapult your business forward or just somebody who refuses to be left behind and want to advance your career, this is the show for you. I'm your host, Isar Maitis, a serial entrepreneur and an AI enthusiast. You'll hear invaluable practical tips from innovative business leaders, AI practitioners, and some of the brightest AI minds in our world today on how you can leverage AI in ethical ways to advance your career and grow your business.
4:15Hi, Isar. Hi, everybody who is listening. Thank you so much for having me. It's a great pleasure. We spoke before. I spoke to your partner several times. I've been following you guys on LinkedIn and trolling you since you guys started posting this kind of stuff. What you guys are doing is magical, right? Because it's really the holy grail of content creation is creating video that used to take, again, a very long time. And now the things you're doing and that the people that are using your tools and your processes are doing is nothing short of magic. Like in a matter of hours, you can create a 10 minute video that tells an entire story about something and it looks amazing.
4:55And it's all because of this technique. So I really leave the stage to you to lead us through the process. How does this thing work? What are the tools? What are the process? So people can learn how to do this as well. Cool. Great. So my presentation is quite short, but I think that designers and non-designers and business owners can benefit from this in various ways. You don't have to be a musician to enjoy listening to the music. Same way I'm thinking about creating stories. You can create stories and tell your story with the characters. So I think that will be a great session and looking forward to sharing my best tips and tricks with everyone.
5:39Let me start by sharing the famous quote that I personally really love and connect with. This is the quote from famous Seth Godin, the god of marketing, someone might say, that AI is the biggest change in our world since the invention of electricity, a bigger change than the internet. And I couldn't agree more. On this slide, if you are watching this, you may see how AI illustration is evolving from 2022 to 2024. This is just amazing. We can create super photorealistic images, mood boards, and now we can create AI illustrations. All of those images that you see right now on the screen are created with a tool called Me Journey.
6:25This is the AI image generator, which is one of the most popular right now. This is the leader of the market. And you can create amazing characters, human, animals, whatever you can imagine. I want to pause you just for just one second. For those of you just listening and not watching this, you're missing like the cutest board you've ever seen because it has 10 different characters that are like a bunny and a little girl and a little girl dressed like an Indian and two little twins and a furry cat. And they all look like something that Pixar would generate, right? It's just these cute, big-eyed, furry, fluffy, cute people or animals, and they just look amazing.
7:06And you just want to hug each and every one of them. Yeah, that's true. That's true. One of my favorites. Those ones I have created over the last year with MeJourney and just wanted to share what is possible right now. So while you can create all of those amazing illustrations, the high detail, the just the texture the quality it's really challenging to create the same character in different poses or different expressions to tell the whole story so consistency with ai is challenging and what we were listening that people are going through the early adopters who started using ai that this is like the biggest challenge.
7:53So one year ago, after many calls with our students, with our clients, we have already tested everything what is possible within the existing workflows. And we realized that there should be a better solution. And this is how we come up with the idea to use the workflow, which we are usually using for creating characters and create a consistent character GPT. It was created by my co-founder, Sachin Kamas. In case anyone is not following him yet, I highly recommend doing so on LinkedIn. He's sharing amazing... And I will second that. Definitely follow Diana and Sachin on LinkedIn if you're interested in anything that has to do with graphic design with AI.
8:36I think so much, Isar. So he had developed this consistent character, GPT. And until now, half a million users have tried using and creating their characters with this solution, telling their stories. And it was just an amazing journey learning from their experience, what they are looking for, how they want to convey the story. And I would like to tell that what are the capabilities right now? First of all, you can access ChatGPT, Consistent Character GPT for free. You will have limited credits in case you are not a ChatGPT plus subscriber. Nonetheless, you can test it out even tomorrow. Just go navigate and look for consistent character GPT and you'll find it.
9:23You will be able to create the same character with different expressions, with different action. It can be human or non-human character and with different poses. I do have an amazing example of AI animation created by one of my AI creative Accelerator student, Tali. She is a psychologist. She is very passionate about helping kids overcome different kinds of challenges, and she wants to create educational content. Right now, AI enables her to do so. She had used a consistent character, GPT, to create many images of the same character, and later she had animated those to create that story. In case you are watching this, please enjoy.
10:10Sparkling waves in an ocean full of wonder lives a unique mermaid with a shimmering tail named Mia. In this enchanted ocean, every creature is unique. But sometimes, even in the most beautiful places, we can feel unsure of ourselves. Mia feels sad and frustrated about her tail. She really believes it's too shiny. Why does my tail have to be so shiny? Everyone scares at me. Mia decided to isolate herself from the others. Hey girls, did everyone see Mia today? No, I didn't see her around. Wait, is that Mia behind the coral? Why does she look so sad? I will go check to see if she is okay. Chloe, that is a great idea.
10:46Hey, Mia, are you okay? You seem to be a bit distant today. It's my tail. It's just too shiny. Mia believes that she needs to resemble others to be considered beautiful. I just want to be like everyone else. Oh, Mia, but your tail is what makes you special. It sparkles because it's unique, just like you. I don't know. Sometimes I just don't feel like I belong. Mia, if you look around, you'll see that no two mermaids are the same. This is awesome, again, to tell a little bit of what we're watching. You can listen to the cool voices if you're just listening to this. But what we're seeing is, again, it's a Pixar level kind of animation of mermaids underwater with the shimmering lights through the waves and the fish swimming around and the coral all around them.
11:26and they're speaking and their lips are moving properly as far as syncing to what they're saying. It looks like a real video. And again, the beauty of the work by Thule is that she is a psychiatrist. She's doing this with some very serious logic behind this on exactly what they're saying and why they're saying it. And it's just, again, a way for her to help her connect with children. And so this is a part of her line of work that allows her now, together with the technique that we're going to learn now, how to create these kind of videos, That is absolutely amazing to me. By the way, we're going to have Thule as a guest of the show early next year.
12:01So you will be able to see more of that aspect of it later on. Yes, this is an amazing project that she has done. And I just wanted to mention that she doesn't have a design background. So this is super empowering how you can tell your stories and create those kind of characters to deliver it in a more impactful way. I do have a little short demo of how you can create the characters with Consistent Character GPT, and I'm happy to walk you through the process in case you are listening to this podcast. When you will navigate to Consistent Character GPT, you can click the button and the chat will ask you if you want to create a male or female character.
12:42So you will just answer in the chat male or female. In my case, I have chosen to pick up the male character. So the next question would be, what would be the name of this character? And I have chosen the name Tom. And then the consistent character GPT, just with this information, will ask me to provide more details. What is the age, country, hair, outfit, or any other details that I would like to have in my image of my character? So I'm giving a little bit more information. my character is Tom he is a boy he's 10 years old and he has very messy brown hair he has blue eyes and I'm describing the outfit that he is wearing in just one minute or less the consistent character GPT will create an image for me according to this description which is super cool because I didn't have to go through any difficult process and I am able to already have my first image of the character.
13:44The next chat GPT will ask me do I want to have more expressions or poses for this character and in my case I'm choosing that I would like to create more expressions so I'm creating a happy expression of the tom and next I'm creating the sad expression and next I will create very excited expression and so much so forward whatever my story has to convey I can create it next. The last image that I'm creating here is Tom being sleepy just to showcase the capabilities of this tool. So I highly recommend and encourage everyone who wants to create a content for their social media educational content or just to have a really fun project one evening with their kids.
14:31It's super fun. They should just go and try this tool for free. It's available in ChatGPT store. And as I mentioned, many creatives have found this ChatGPT really useful. You can navigate to YouTube and just type Consisting Character GPT, and you will find a ton of guides, tutorials, and step-by-step process described with the details, how professional creatives are doing this, or just beginners. This is awesome. I want to ask you, oh, you're going to, what you're about to do is exactly my next question. So I will let you continue. But you know what? I'll ask the question just for the audience because we started with, okay, I can create these images with Me Journey or I can create them with whatever other tool.
15:15A lot of people now switch to other tools as well. But if you have an external tool that you're using, or it's a real image, it's an actual person that you have an image of, how can I do the same thing and then bring it into the AI universe, make changes to it and create videos or even just the images themselves in various moods and character positions? That's a great question. And the next possibility that I wanted to talk about to use your own images or even AI-generated images and create more character variations is by using MeJourney. MeJourney is one of the leaders at the market right now and it does have a learning curve but it impresses everyone with the results that you can get.
15:59To create consistent character in mid-journey what you can do is to generate the image and use this image as a reference which means you will upload this image as a part of your prompt and before you will do so you will use a command dash strep which means the character reference. And this is working amazingly with photorealistic results. You can create an experiment with the backgrounds, with the outfit, with different situations and different poses. And many creatives find it specifically useful for creating storyboards for pre-production. It had saved a lot of time. So this is a really helpful tool to use.
16:46And in case you have not experimented with this yet, I highly recommend going and using your own photo or AI generated photo and trying out this command, there's sort of image journey. So just to add to that, for those of you who don't know me journey, first of all, as, as Diana said, it's probably the best tool out there right now together with flux, but the mid journey has been around for longer. And as of about a few months ago, they have a proper user interface on their app. And so even doing this became a lot easier because after you create an image or upload an image on the right side menu, you have a button that says, use this.
17:20And you have different options of what you can use. And one of them is to use the image. And then if you literally click that button, it will do that process for you. So you don't even actually need to know the command. You can click on use the image, and then you will put it in the command line for the next generation. And then you can add whatever you want to add, and it will use it as a reference. The other thing that you can add is you can add how strong you want that reference to be. So do you want it to be to relay on that image a lot or less? And with that, with those two parameters, you can control how much it's going to look just like the original image and how much variation is going to be in that image.
17:54But it's a very powerful capability. Yes, you're right. Although I must tell you that it works amazingly for photorealistic results and a little bit less amazing for illustrations. You can create a character and use this character as a reference. And right now on the screen, you can see the images that I have generated within Midjourney and use the same character as a reference to create a side view or the shot from the behind to generate the same character in different angles. Nonetheless, in case you want to have more control over the pose, it becomes really challenging. while midjourney is creating amazing illustrations detailed high quality with the background and it can convey the story the part with the consistency is still really tough to figure out so it will take you a lot of prompting a lot of time to generate many variations and then to pick from those which ones look the most similar you don't have so much control as you would really want to have and for beginners it might be a really difficult path to go through.
19:08Right now on the screen I'm sharing step by step how you can do this in MeChurney. As a step one you are creating the character, as a step two you are adding this as a character reference and you're modifying your prompt a little bit to create a different angle view and also you can use the same image as a style reference to ensure that the style is the same. Although this is possible, but it's really challenging. Another way to create the same character in different poses from different angles, and many game studios is using this one, is prompting character sheet in your prompt while creating your character.
19:49And if you're watching this right now on your left side, you can see the results generated with me journey by using character sheet in the prompt and you can see that you really need a lot of generations to have a good result. You have the same character from different angles but something is missing there or the pose is just repetitive and you don't have so much control. When you will generate 20 different variations you will pick the one where you have the same character from different angles but you will not be able to control the details of the outfit that much or of the character itself. And especially later with the poses, you will not be able to generate something new.
20:33So just to summarize quickly and to give people who are listening what we see is when you use the concept of character sheet, it gives you several different variations of the same character. So side view, front view, 45 degree view, but you have no control on what that's going to be. So yes, it gives you multiple variations, but if you want to have them in specific poses of variations. You just have to generate the same thing 25 times. And then out of those, grab the three. And in each and every one of them, you're going to have a few of the characters, right? So in the different setups. So if you run it 25 times, you're going to have between 75 to 100 options of this character.
21:08And it's just a lot of work to get the... And then you'll have to pick the one that probably is not exactly what you wanted, but is the closest to what you wanted. And out of 100 options, you'll have three or four that will be roughly what you need. But it's just time consuming. That's right. It is time consuming and agencies are picking the closest version. Yes. What my journey is great with is that you can work with really different styles, but it's really challenging to get control. And later when you want to generate the character with specific posts, it will not be possible. What you are looking for, you can see right now on the screen on your right side, where you have the character, you have the same exactly outfit all the time from different angles, from different views.
21:50And later you will be able to generate more poses or expressions for this character. Not similar one, not after 300 times that you try, but very efficiently and very fast. So I am calling me journey lottery because of amount of times you generate and depending on the character and the outfit that you want to create, It can be really easy or really tough. I was creating one project for my client and it was so challenging. I have spent the whole day for another project I was generating the character. It was possible within one hour. So it's a lottery, a lot. You have to be lucky and you have to be lucky specifically in this particular project.
22:39I'll pause it just for one second. And my son really likes manga. And we were trying to create a manga for him that he, him together with me and Claude has written the storyline, but he had the general idea. And then we did exactly this, right? We were trying to create the images, like the manga images for the story. And it was ridiculous because it's a little ninja with a stick. And then sometimes instead of a stick, it has a sword. Sometimes his ribbon on his head is red. Sometimes he doesn't have the ribbon as a whole. Sometimes he has a hat. I'm like, what the hell? I'm using the character reference.
23:08Like you're supposed to show me the same character. And it's just, like you're saying, it works, but at X percentage of the time, but then it's not exactly the angle. And the other guy doesn't look the same. And the tower behind him doesn't look the same. And so really getting to a situation where you can consistently create what you want without spending hours of recreating stuff and picking is the holy grail. Yes, that's true. And also it's true for beginners because it's really challenging for them to create consistent character as much as for creative professionals who value their time and they need to do this at specific deadlines.
23:44So they have heavily to rely on Photoshop in those results to have more consistency or something like that. So let's talk how it is possible to get consistency. After we have seen the success of consistent character GPT, and at the same time we gathered a lot of feedback, what can be improved, we have decided to partner with the SegMind. This is the platform where we can deploy the workflow that we have developed. And with this workflow, it will be possible to generate a character sheet in less than one minute. Super easy, super fast. and right now on your screen you can see the character sheet which has the same character in full body pose as well as just the avatar and that gives a lot of possibilities to already tell the story or use those characters for some book or illustration or for presentation as it is without specific modification.
24:43So how to use this? I will probably share all the important links with the ISAR. But just to give you the roadmap, you can use the QR code, which you can see right now on the screen. You can scan it and that will open a guide. In this guide, you will have just two links. First link you will log in with and that way you will get free credits. And second link is specifically this kind of workflow where you can type the prompt and get your character created. I have a specific prompt structure that I would highly recommend in case you are creating characters and you want to create consistent characters in your future generations.
25:26It starts with image style, because this is important. Everyone works with a different one. Character type, where you describe your character in more details. Then you have age, hairstyle, outfit, emotions. And usually I do recommend using a clean background. because you can very easily create the background later and overlap them in free tools as well. So the example prompt would be Pixar style, French girl, 10 years old, curly brown hair, simple red dress, happy smile, clear background. And I would highly recommend everyone who wants to create consistent characters to use very simple kind of outfits because AI can hallucinate and you will have more challenges along the way if you pick something super specific, very detailed and overcomplicated.
26:16I do have a short demo video how you can create this. Here I'm showing the SACMind and you can just upload your prompt following the prompt structure which we just discussed. And in less than one minute, the model will generate the character for you. Are there any questions, Cesar? Well, it's - Yeah, so first of all, it's a tool that you guys created, right? That is now in beta and it's going to be released sometime soon to the world? This is the workflow which we have deployed in partnership with the SegMind. It's already available. It's not the app yet. The app has some extra capabilities. Got it.
26:56Okay, so this is available. Anybody can use it. We'll put the link in the show notes. So if you guys want to try it out, just open your phone right now or wherever it is that you're listening to the podcast. If it's on your screen and then just click the link to this and you'll be able to create these consistent characters very, very quickly. And by the way, as we're looking at this now, we already have the girl with her address and the curly hair showing up. And to tell you what it creates, four images of the girl from front, side, diagonal and back. And then it creates a similar thing just of her kind of like headshot, like just the avatar of the same girl.
27:28And she looks exactly the same and the dress looks the same. And she has the same kind of pose in all images and she has the same shoes in all images, even if we didn't prompt that. And so it's amazing how good this is compared to, if I'll use Diana's words, the lottery that you have to play X number of times. Again, because it's the lottery, it might be 20 times, it might be 200 until you get what you need. Yes, and you definitely cannot just do it in less than one minute in my journey with this kind of results. Quick question about the costs. You said there are going to be a few free credits or tokens.
Read the full transcript
28:02What's the cost if you want to start using it regularly? Like to create a sheet like this, how much would it cost somebody? I think it's just a few cents. Oh, okay. So it's not... Yes, because it's using the power to create those images. The prices are put by the SegMind and you can read more about this when you're logging in. But it's really cheap. There is no huge investment. Definitely worth paying a few cents rather than spending 20 minutes to an hour. So I... That's true. For sure. and you can create your characters with different nationalities ethnicities as well as different age and different style right now on the screen we can see the very cute young girl and the very old grandpa and sometimes it's challenging with another ai image generators to have this kind of consistency for different ethnicities and this is a great solution for this what you can do just oh So again, I think I was about to ask exactly the perfect question because what you're not seeing, if you're just listening to this, is that they have a clear background, as the prompt said, which means it's very easy to cut them out and then place them in whatever background you want to generate with another tool, which is exactly the next slide from Diana.
29:13So I was about to ask the right question twice. So I'm giving myself a little credit here. You're on the track. So yes, what you can do next is just to use this character sheet. I would recommend Convo. It's free and it's very easy and simple to use. And you can cut those images and replace the background directly in Canva with its AI capabilities. Or in case you're using any other image generator like Dali, Ideogram, MeCherny, whatever you prefer, you can generate just the background and then overlap that together. That way you will have the beautiful view which tells the story and your character.
29:53and in the next scene, you can create something else and use another pose. What else is possible? Can you believe that my character was created in just 40 seconds with consistent character AI? I think this is amazing. You can use the avatar and cut the simple avatar and animate it with another AI tools. For example, in this case, I'm animating the avatar using Heijen. I can also use HeyGen for free in Canva. Super easy, super simple, and very user-friendly. You can upload the text that you want your avatar to say, and it's creating the lip sync for you. You can choose the tone of voice that you would like to use for your character and have such amazing results.
30:41In case you are curious what other tools you want to use for animating your characters, I would highly recommend Kling. I would recommend Runway. Those are one of the biggest players right now and they create amazing results and as well you can use the first and the last frame of two images to animate and create very smooth kind of animations for your characters. So to again touch a little deeper on what you just said, HeyGen is a tool that allows you to create avatar videos and these could be made up avatars you can create yourself as an avatar, you can bring it in and then you can put in text, it can translate it to any language and he will speak it in any voice out of the hundreds of voices they have or your voice if you're training to that.
31:26And you can use Heijen within Canva to save yourself a few steps. So you can get the images that we created with a consistent character, bring it into Canva, remove the background with one click. There's literally a remove background button. So the base background that was created with the sheet disappears. Then you can create your own background in there or bring it from somewhere else. And then you can animate the avatar using the Hagen engine within Canva. So that puts that all together. The other tools allow you to do a lot more so that the Kling and Runway allow you to create a whole video.
31:57And you can start either just by text or you can start with text and an image, which is what we're suggesting to do here. So you can create the avatar in the consistent avatar. You can bring a background and put it together in Canva and then take that as the initial image in your video in Runway or Kling. but what Diana said in addition to that you can take a first image and last image so let's say what you want to show is you want to show your character walking down the street and then turning a little bit if you put the first character further away in the street facing you as the first image and then the last image as the same character facing sideways closer to the screen the video generator in Kling or Runway as of a week and a half ago because it was available in Kling a while back and Runway just added it, will understand that, okay, I have this, the background stays the same.
32:47This person is over here. Now it's over here and it's facing this way. And it will create a natural movement for that character walking and making that turn that you intended. And if you think about it, now you can stack a few of those together and create a full video from different angles, from different scenes, where every segment is five seconds, six seconds, 10 seconds, depending on the tool that you're using. But then you can edit them all together into a flow of a video that will show the full story. Yes, that's right. And also, in case you're looking for a free solution to test animation, you can use Luma.
33:21And it also has the capability of first frame and the last frame, but the generation can take quite a long time, up to 10 hours, so maybe longer depending on how many users are using the app at that particular moment. So just keep in mind that it can take a while. Nonetheless, it's also great AI animation tool. So in case you're not familiar, check it out. And here we are coming to the solution that I am super excited to talk about because it's consistent character AI app, which we together with Sachin Kamath had co-developed and launched very quietly just recently for our small group of beta users.
34:04Oh, you heard it here first. That's true. That's true. we really are looking forward to getting more feedback and react and improve the user interface and their experience and as we are bootstrapped we are doing everything just two of us we are launching this in very soft launches in very small groups so in case anyone is seeing the potential in this what i will talk about just feel free to reach out and maybe you'll get the access a little bit faster, but you can also join the wait list and get access when the luck comes. Yes. So what does Consistent Character AI can do for you? First of all, it will create a consistent character sheet as we had seen before.
34:51You will have the same character from different angles, and later you will be able to create it in different poses. Second of all, you can upload the image reference even in case you're struggling with the prompts and you don't want to spend your time on prompt engineering. You can just upload the image and automatically within 30 seconds it will write down the prompt for you within the application. Third of all, you have a very user-friendly dashboard and you can come back and create the same character in future for a different story or you want to experiment with more poses for example it's possible also we will have super exciting additional features like chat gpt within the application and one more secret feature which is called turbo i love it it looks like a chat and you can create not only human but non-human characters maybe the quality of consistency is not there at such high level as for human characters Nonetheless, it creates the illustration super fast, and it's an amazing solution for brainstorming or pre-production for agencies or for creatives.
36:05And I'm happy to showcase what is possible right now, at the moment. After one week, maybe it will be better. So you can create the same character with different poses, and it will be super consistent through whole period. Later, you will create separately your background, and you can tell your story like we have discussed before. We already know, and we have tested this, that consistent character AI can reduce pre-production effort up to 80%. And depending on the use case, it can be even higher. So I believe that this is a great tool, and it has proven that our clients that we are working with, they have created amazing storyboards, game designs, and they are super passionate about continuing exploring this solution and creating more characters with it.
36:57This is how the dashboard will look like. And the promise is here, full control over character process. This we can deliver. In case you are looking for this kind of solution, this is the best what I know at the market of AI design right now. I want to pause you just for one second. So we talked a lot about character consistency. And really the only thing that's missing right now, which this app solves, in addition to solving all the other problems all in one place, is pose control, right? So you can create characters, but now you want them in a specific exact pose, leaning on the desk, looking dreamy as it's lying in bed, jumping up in the air, waving her arms, like whatever the case that you want to do, dancing and twirling.
37:38It's something, again, you can do right now with the same lottery that we talked about before. And then eventually you might get what you want versus here, you can control the pose. And I've seen the early demos that you guys have shared online. You can control the pose either with a prompt or actually graphically controlling the pose, which is even more amazing. You can literally grab the face and move it around so it looks in exactly the right direction or move the arms around and so on. And so I see this thing as extremely exciting. Thanks so much. And we are looking on developing new features.
38:11So we are also excited about this. Right now, you can create your characters in five different styles, which are 3D cartoon to the illustration, photo, anime, and Pixar style. And here is what I have mentioned before. On your screen you can see the dashboard where you can upload the image reference, your existing image, and it will automatically create a prompt for you to illustrate that image and it will follow the specific prompt structure that works the best for this solution at the moment. And you can name your character to make sure you can come back to the dashboard and find it later. You can create the same character in different styles.
38:51And I believe that this is the best time to dive into the demo. In case we have any questions during the demo is playing on the screen, I'm happy to answer those because I have included the live demo. I did not want to speed up Awesome. And there is why I just wanted to show the real thing here. So I'm happy to cover any questions in the meantime. And here I just showcase the reference pose and my character, what it will be. So let's see how to create such. First of all, we are navigating on the consistent character AI to create new character. We are typing the name of the character and we are choosing the style.
39:34In this case, I am not using any image reference and I'm just waiting for my character to be generated. We will have the four different variations of the same character with different angles. And we already created them not only in one character sheet that you don't have to cut and worry about this, but each pose and each angle is already in separate image. What can you do now? You can press the button, generate new posts, and just by using the image reference of the posts that you would like to generate, you can upload it directly to the dashboard. The system already will use the character that you have created before as a reference and the reference posts that you have uploaded.
40:25and in less than one or two minutes, depending also on how many users are using the application, you will have the results. And you can - I just want to pause you again to share with people who are listening. We, when I say we, as if I have any part in this, Diana and her amazing tool has created a character. It's a girl, she wears jeans and like a square pattern shirt that she's wearing. And she was standing in a regular pose. As we've seen before, four images, one looking forward, one looking 45 degrees, one looking backwards, and so on. So four images. And she uploaded an image of a kid ninja leaning backwards.
41:02And then as soon as the thing was running and as we're taking, it's already there, it created the girl with the jeans and the sneakers in the same exact pose as the ninja guy. So you can, I assume, take pictures of yourself in the exact pose that you want. So you don't have to prompt anything. Or you can take any picture of the internet that is exactly what you want. And there's a lot of these and use that and it will then create your character in exactly that pose, which as we mentioned before, is not doable right now, literally in any other tool that I know of. And you can use, as Isara mentioned correctly, any image reference.
41:41What I wanted to share with you that the best what can work is the square one-to-one aspect ratio pose reference. And you can easily change the aspect ratio also in Canva. Maybe in future we'll have the feature which will do it automatically for you. And that will ensure that proportions when you're uploading the image reference will stay the same throughout the character generations. But look how simple it is. You already have within one minute your character. And now in less than a few seconds, you have the same character with different kind of poses. The character has the same face, the same hair, the same outfit.
42:20In case, sometimes depending on the outfit, it can change a little bit and we cannot control that at such high level. But for even played shirt, it works perfectly well. Super consistent. I can tell the story. I can see also that emotions a little bit are changing and I was not using any prompt whatsoever. So over what I was using is the image reference of the post, which I can just find in the internet or make a photo of myself and upload it. Wow. Okay. I want to do a quick summary and then I'll let people know where they can find and follow you. But we started talking about the importance of storytelling, right?
42:58That's what this is all about, right? You want to tell a story and you want to tell a story either with an image or with a video. And you can now do this with AI in amazing ways, other than the fact it's not consistent. It's very hard to create. the same character. And if you want to tell a story, then you have to have the same character. And Tiana and Sashin has created is the ability to do this in multiple different steps. You can use their GPT for free, unlimited, as long as you have a paid GPT account. If you don't have a paid GPT account, please get a paid GPT account. Like chat GPT does so many things for 20 bucks a month, like no tool ever in history gave you so much for so very little money.
43:33But I'm putting that aside for a second. So that's option number one. Option number two is she gave us the tool to create these character sheets that is for very little money that is available right now. So if you want to tell a story, either it's something you're doing for yourself for fun, something you want to do with your kids, I highly recommend that. It's an amazing activity with your kids to teach them about AI because they can use ChatGPT or Claude to write the story and then the scenes, and then you can use that tool to create the actual characters. And then the final thing is the holy grail, which is the post control combined with all the other stuff that is coming up in an app that they've been developing.
44:02This solves so many problems, whether you're an agency that really needs to do this on the day-to-day, or you're doing this now, maybe, I'm not going to say that out loud, but I am, maybe you don't need an agency because now you can do a lot more stuff on your own. So if you're a small business and you cannot afford to hire an agency, you can now do these things on your own with either yourself or your marketing team in-house in something that was literally impossible to do before. You can now do on your own. And once you figure it out and learn the process, you can do it in minutes. We literally just did this live and within, I don't know, three minutes, we had the same girl in multiple poses.
44:36So you can tell your story. Absolutely amazing. If people want to follow you, work with you, learn more about what you do, what are the best ways to do that, Dan? The best way is to connect with me on Linden. But before we will dive and I will share with you just the simple QR code, which you can scan and connect in one clip. I just wanted to add a little bit that in case you want to create your characters either they are photorealistic or they are cartoonish style it is possible and it's super easy and although you also might generate your character a few times or regenerate a specific pose it had it had been a wonderful solution and we are getting very positive feedback from our clients and our students who are testing this application right now and i'm just getting ready for today our live session in our AI Creative Accelerator where we are teaching to create various kind of AI designs with various kind of AI tools.
45:36So I would highly recommend everyone to test out this solution. In case you are wondering how to use MeJourney or in case you are wondering how to use ChatGPT for free, please feel free to reach out to me as well. I'm happy to help with whatever experience that I'm having and help everyone navigate through this time. And here is the QR code, how you can scan and super easily connect with me. In case you're listening, just type in LinkedIn, Diana Zdibel, and you will definitely find me. I'm redheaded1. Let's connect and learn together. Awesome. Diana, thank you so much. This was absolutely fantastic.
46:19And I think people, any people with any creative ideas in their head, whether they're working on an agency or just they want to do this for fun, I know we'll use this and we'll find this really helpful. I really appreciate you taking the time and sharing with us. I really appreciate what you and Sachin are doing for the AI community and the creator community. So thank you. Thank you so much. Thank you so much for having me. In case anyone is curious, feel free to join the waitlist for Consistent Character AI and you can test it out on your own. it had been a great pleasure being on this podcast and sharing my tips and tricks with you and thank you so much again
From the publisher
How do you tell the same story with consistency across hundreds of visuals?
Storytelling has evolved from spoken tales to AI-generated videos and photorealistic characters—but one challenge remains: consistency. It’s easy to generate impressive individual frames with tools like Midjourney, but what if you need to maintain the same character across multiple scenes, moods, and poses?
In this episode, we welcome Diana Zdybel, co-founder of Consistent Character AI and Neolemon, to break down the art of consistent character creation with AI. If you've ever felt stuck generating high-quality visuals that align perfectly with your business brand or internal messaging, this session will change your game.
Through tools like Consistent Character GPT, Diana shares how to maintain visual harmony across AI-generated characters and illustrations. Whether you're an entrepreneur building pitch decks or a marketing team creating ads, Diana’s insights will help you tell compelling, visual stories—efficiently and at scale.
In this session, you’ll discover:
- How AI revolutionizes storytelling for business communication.
- Why consistency in visuals matters for internal and external messaging.
- Practical tips for creating characters with Consistent Character AI.
- How tools like Midjourney and SegMind can streamline your workflow.
- A walkthrough of the most common challenges in AI-generated visuals—and how to overcome them.
- Bonus tools and tricks to animate avatars and characters using HeyGen and Canva.
With a background in AI design and years of experience navigating the challenges of character creation, Diana helps businesses streamline their creative workflows. Connect with her on LinkedIn to stay updated on her latest projects: Diana Zdybel.
Find more about Consistent Character AI here: https://linktr.ee/neolemon
About Leveraging AI
- The Ultimate AI Course for Business People: https://multiplai.ai/ai-course/
- YouTube Full Episodes: https://www.youtube.com/@Multiplai_AI/
- Connect with Isar Meitis: https://www.linkedin.com/in/isarmeitis/
- Join our Live Sessions, AI Hangouts and newsletter: https://services.multiplai.ai/events
If you’ve enjoyed or benefited from some of the insights of this episode, leave us a five-star review on your favorite podcast platform, and let us know what you learned, found helpful, or liked most about this show!



