In this interview, Niklas Vesely speaks with Adeh DeSandies, co-founder and CEO of InstantStudio AI, to explore how AI-powered image generation can be seamlessly integrated into websites and existing business projects. Adeh discusses the simplicity of adding their software via API, the wide range of applications from professional LinkedIn profile photos to custom avatars in gaming, and the cost-effective solutions available for businesses looking to enhance their digital experience. He also highlights success stories, InstantStudio’s expertise in portrait generation, and the team’s journey into the AI industry. Whether you’re an e-commerce director, a gaming developer, or a marketer looking for innovative tools, this interview dives into the practicalities of AI image generation and how it can be scaled for any business.
Niklas Vesely:
Hey Adeh, good meeting you. I have a few questions I’d like to go through with you. I’d like to start with some simple questions on how to actually implement or add image generation software to existing projects or existing websites, then maybe talk a bit about the team, and end with the most pressing question most people have—how much does this actually cost? But let’s go step by step. So, let’s say there’s someone who’s got a website and they could really make use of image generation software that also edits photos. How hard is it to put your software into a website?
Adeh de Sandies:
That’s a great question. It’s not very hard to put our software into a website. What we really specialize in is providing solutions for businesses that need AI tools but maybe don’t have the expertise to do the research, get the GPU access, or set up the private cloud. Since we already have our private cloud set up and accessible via API, there are many ways we can do an integration. We can write the software and integrate it into your site, or you can write the code that talks to our API. Depending on how you want to handle it, you’d either pay us for the integration or handle it yourself. But, given our high-level APIs, you could add functionality to a website in less than a week.
Niklas Vesely:
I have some follow-up questions, but I should maybe go one step back. What kind of image generation software can you make? Are there limitations? For example, a user profile photo upload should look professional, but that would be completely different from someone uploading a photo of their pet or even products like clothes. What are your specific areas of focus, or can you do everything?
Adeh de Sandies:
That’s another great question. The applications for image generation are vast. There’s an endless variety of images that can be generated. Technically, we use open models like Flux and Stable Diffusion to create custom solutions for specific problems. If you want to upgrade a profile photo to a professional LinkedIn-style image, that’s something we manage well, and we already have an API for that. If you want to change the style, say for a gaming website where users create their own avatars, that’s another popular application. We also do things like ingesting artwork or marketing assets from companies and training the AI model to produce images that align with the brand’s colors and style. It all depends on the client’s needs, and we work closely with them to deliver what they require.
Niklas Vesely:
Now I heard some terms like prompting and profile photos for LinkedIn. How do those come together? What does the actual process look like, and what’s included in such a project?
Adeh de Sandies:
Good question. I can walk you through a customer success story. We provide what we call the GPU Compute Cloud—basically, a private cloud with Stable Diffusion and Flux API GPUs, all accessible through APIs. What this means is that we privately handle all your data. Nothing goes to third-party providers, and we delete the data afterward. When a client comes to us, for example, wanting to offer free professional-style photos for resumes, we provide the API call that handles that. The inputs would typically be a selfie or a series of selfies. Then we decide if their website links to ours or calls the API directly. We quote the project based on the estimated work, implement and test it, and afterward, the client tests it on their end. From there, there’s a simple usage fee based on server usage—so you don’t get charged excessively if it’s not heavily used.
Niklas Vesely:
Cool, and then, like other AIs on the internet, I know you can upload a photo and say, “Make me look like a dragon,” for instance. Do users of your software write those texts themselves, or how does that work?
Adeh de Sandies:
That’s a great question. Prompting is essential in text-to-image generation. While text-to-image tech is available in places like ChatGPT and Google, our focus is on simplifying the process and providing a simple user interface. Instead of the end-user typing in prompts like “Make me look like a dragon,” the client would define those prompts on the backend. For example, you could specify that when someone uploads a photo, it turns them into a dragon. We work with clients to design those prompts and backgrounds and ensure the solution is seamless, so the end-user just uploads their picture.
Niklas Vesely:
Perfect! So, going back to the LinkedIn example, a job board could have users upload a selfie, and their profile would show a professional-looking suit picture, right?
Adeh de Sandies:
Yeah, yeah, I mean that’s the idea! There are limitations, of course. For instance, if you upload a picture of yourself looking sideways or sticking your tongue out, you might end up with a picture of you in a suit still sticking your tongue out. So, there is some give and take—it’s not magic. But in general, we can definitely give you a very professional-looking, dressed-up picture without you having to dress up, put on makeup, or go to the mall for a photo. It saves everyone a lot of time.
Niklas Vesely:
Okay, oh, this sounds great! Then, let’s say a customer now says, “Cool, this kind of software would be amazing on my website, like a job board.” What does the whole process look like? Are they looking at one year of really hard, intense work, or is it like a 2-hour thing? What are the steps in between?
Adeh de Sandies:
Yeah, I mean, that is a great question because no one really knows. Everything is so new, and everything’s changing so fast. What I can say now is that we do have certain specialties. One of our specialties right now is portraits. So, if the functionality you want to add is related to portraits, it can be very fast. We work together on designing the portrait, designing the theme, and how it looks, coming up with the prompts, testing it on a testbed of images, and making sure the client is happy with it. That can all happen within a few days, and the software is pretty much ready to go. We just need to do an integration, and everything can be done within two weeks. If the application is something more involved or involves different technologies like video or other things that require more preparation, it might take a little bit longer. But because we have our scalable private GPU Cloud and API access, for the most part, you’re looking at no more than a month to get an application up and running.
Niklas Vesely:
Okay, nice. Just a quick follow-up there—you mentioned the word “video.” Bless you, by the way. You mentioned the word “video”; do you also do videos?
Adeh de Sandies:
We can. We have the capability to produce 4- to 6-second videos based on images, sort of like a “live photo” in iOS. That’s where we’re starting, and we do have plans to build on top of that in the future. But at the moment, that’s where we are.
Niklas Vesely:
Nice! Okay, so we have the question of how long it takes—around two weeks for a project—and then, like two weeks for an AI project seems quite fast. So, who’s the team behind it? Who are the people that make this happen, that you can create custom image generation software at this kind of speed?
Adeh de Sandies:
Sure. Right now, our team consists of a bunch of people, most of whom are software developers who have been working with me for over 16 years. We’ve been working together for a long time on various software projects. We also have AI researchers and a content team. The content team is dedicated to producing ideas and generating the prompts that come up with the good-looking images we create. Together, the team works closely to build great software with great applications at a great price.
Niklas Vesely:
Cool! You mentioned you’ve been working with your team for 16 years, and you have various projects. Are there some people who stand out on the team? Who are they, and what are some of the big projects you’ve worked on that brought you all to this point?
Adeh de Sandies:
Yeah, they’re all sitting here with me. We have Shaun over there, he’s a software engineer. We’ve been working together since 2006. Then, there’s Kenny from Lithuania, he has a master’s degree from Lithuania and has been doing software engineering with us since 2008. My co-founder, Oleg G., is from Germany and worked in Silicon Valley for over 20 years. So, we have a very diverse team working together to get things done.
Niklas Vesely:
What’s one of the biggest projects you’ve worked on together as a team?
Adeh de Sandies:
We did a lot of what we call high-stakes auctions, where we developed and operated auction systems for very high-value things. One of the projects we worked on was the auctions for the “.com” domain name endings. So, if you’ve heard of things like “.photography,” “.restaurant,” “.cafe,” “.LLC,” “.doctor,” “.lawyer,” all of those. Around 2015 or 2013, we operated an auction that resolved an issue they were having. We took a bunch of money that had been dedicated to launching these domain names and distributed it among the people who were competing for the right to use those domain names. It created a lot of value for many people, and we’re really proud of that project. We also helped the Thai government sell almost a billion dollars worth of radio spectrum, which was another great project we’re proud of.
Niklas Vesely:
Oh, thousands of questions come to mind when I hear about these projects, but let’s stick to InstantStudio.app and image generation for now. Maybe a quick self-introduction—what’s your background, Adeh? We’ve been talking about your team, but not yet about yourself.
Adeh de Sandies:
Sure. My background is in computer science—I graduated in 1999, right in the middle of the dot-com boom, the first one. It was a great time to be in Silicon Valley. I was living and working there, and that’s where I met my co-founder.
Niklas Vesely:
Which university did you graduate from?
Adeh de Sandies:
I graduated from Stanford. After that, I went into the industry, doing software development and engineering for a long time. For the past 15-16 years, I’ve been building this team into a highly functioning, advanced software engineering team. That’s allowed us to get involved in some very interesting projects, and now we’re focused on building AI applications.
Niklas Vesely:
What led you to start making image generation software?
Adeh de Sandies:
There were lots of reasons, but two main ones stand out. First, we saw the trend happening about a year and a half ago. AI isn’t going away, and it’s not going to be regulated out of existence. Even as rudimentary as it was back then, AI was already starting to show its usefulness. Companies like MidJourney launched and gained traction quickly. So, instead of ignoring it, we decided to embrace the challenges. We needed to learn how the technology worked—machine learning, training, data science—and we dove into that. I also have a personal love for photography, one of my longest hobbies since I was a teenager. So, I was naturally drawn to image generation. With some engineering ingenuity and strong software development, we realized we could deliver something cost-efficient and valuable to customers.
Niklas Vesely:
Nice! I think a lot of people listening to this right now are really curious: how much does this cost? Are they looking at a fortune, or is it actually affordable? For example, if a gaming company came to you and said, “Hey, we’ve got this amazing computer game, and we want to use our users’ faces inside the game. We need to adjust the avatars to match the game’s style,” how big is a project like that, and how much would it cost?
Adeh de Sandies:
A project like that would actually be very affordable. As I mentioned earlier, we specialized in portraits early on, and that’s where we’ve developed our software. Something like that would fall under our InstantStudio portraits product. We could get it up and running very quickly—about two weeks of development time. After that, there’s a monthly fee for access to the API and private cloud, plus a usage fee if the usage goes above a certain level. Development would cost around $3,450 per week, and the server would be about $650 a month. We’d need to assess the usage, but overall, it’s a very affordable process that brings a lot of value to the final product. We’re also open to running pilot programs with clients to find new ways to apply this technology.
Niklas Vesely:
Perfect! At the bottom of the video, we’ll add a link to contact you. If people have questions, they can reach out via the contact form.
Adeh de Sandies:
Nice!
Contact form: scroll down to the bottom of the page at Kudestudio.com
Leave a Reply