Subscribe to AIMinds Newsletter
Stay up-to-date with the latest AI Apps and cutting-edge AI news.
Subscribe
Share this guide

About this episode

“I was making these songs, and I knew I had to promote them on social media, and I liked making music, but I didn't like recording myself. I didn't feel comfortable, so I started making lyric videos, and I had content to put on TikTok, to put on Instagram reels and stuff. Problem was that it was a very tedious, boring job, and I thought, okay, I can build stuff, so why not build an automatic lyric generator?”

— Ulysse Maes

Ulysse Maes, the Founder and CEO of Scribewave, is a 24-year-old PhD student at imec-SMIT-VUB in Brussels, Belgium. During his time studying business engineering in Leuven, as a side project he started building custom websites as a freelance software developer. It was through this experience that he discovered his passion for blending academic insight with practical development skills and an entrepreneurial mindset.

In May 2023, Ulysse successfully launched Scribewave, an automatic transcription SaaS designed for journalists, researchers, and content creators. The platform features a specialized, time-synced editor and offers various export options, making a difference on how audio content is managed and utilized. Ulysse's vision for Scribewave extends beyond its current capabilities, aiming to transform it into the ultimate multimodal conversion suite. For instance, the service already excels at converting podcasts into visually engaging visualizations.

When he's not caught up in academic papers or coding sessions, Ulysse enjoys expressing his creativity through Dutch tunes as "MC Ulies, loves a run outdoors, or binging "The Office."

Listen to the episode on Spotify, Apple Podcast, Podcast addicts, Castbox. You can also watch this episode on YouTube.

In this episode, Ulysse takes us through his unique foray into the world of AI and machine learning, starting from his roots as a musician to becoming the creator of an AI-driven, automated transcription service. The conversation ranges from the inception of Scribewave to its evolution and vision for the future.

Here are some highlights from the episode:

  • Find out how his musical career led to the creation of Scribewave as an automatic lyric generator.

  • Discover how experiments with speech-to-text automation and subsequent user requests shaped Scribewave's journey.

  • Listen to Ulysse's key learnings from his initial scalability mistakes and how he turned them into advantages.

  • Understand how Ulysse moved from bootstrapping alone to learning the importance of being part of a team and partnering with others.

  • Learn about Ulysse's experience launching and growing his product, even in a risk-averse environment.

Fun Fact: The initial motivation for creating a speech-to-text tool came from his sister's need to transcribe her interviews for her thesis.

Show Notes:

00:00 Bridging engineering and management for success.
03:27 Freelance software developer turns to entrepreneurial projects.
07:37 Create lyric videos.
11:06 Serverless project.
15:30 Unexpected loyalty of student customers.
19:14 Graduated, doing a full-time PhD.
20:46 Inquiry about growth and progress.
26:13 Outsourcing models to Deepgram, optimizing AWS environment.
29:23 Product success despite mistakes, inspiring innovation and growth.
30:40 Wrap up

More Quotes from Ulysse:

“It's basically you're learning to become the bridge between the engineers in an organization and the upper management. Many of those business engineers aim for CEO positions in a later stage of their career.”

— Ulysse Maes on the bridge between engineers and management

“In the beginning, there was no speech to text automation. It was just me pressing the spacebar every time a word was said, which was really terrible, but was already better than doing it in premiere pro.”

— Ulysse Maes on the creation of Scribewave

“I just looked for transcribe or transcription in Twitter and I checked some posts about it. And then I found a journalist who was complaining that no decent transcription solutions existed for the Dutch language.”

— Ulysse Maes on researching about transcription service

Transcript:

Demetrios:

Welcome to the AI Minds podcast. This is a podcast where we explore the companies of tomorrow being built. AI first, I am your host, Demetrios, and this episode is brought to you by Deepgram, the number one speech to text and text to speech API on the Internet. Trusted by the world's top conversational AI leaders, startups and enterprises like Spotify, Twilio, NASA, the one that sends rockets into space, and Citibank, the one that gives you all kinds of places to store your cash. Today we're talking with Ulysse, who is the founder and CEO of Scribewave. It's great to have you on here, man. I'm excited for our conversation.

Ulysse Maes:

Yeah, thanks for having me.

Demetrios:

So we've been chatting a bit beforehand and I'm glad that we finally were able to connect and record this episode because for those who are listening, it's been a little bit of a roundabout with me having Internet issues and then video issues, and it was all my fault. But you were persistent and you were determined. I get the feeling you are a determined builder also because you are a solopreneur in this space creating scribewave. I want to hear a little bit about your history. What made you become such a determined person?

Ulysse Maes:

Yeah, I guess you could say that. I'm also very happy that you use the word builder because that's really what I identify as. I don't really know a lot about machine learning, for example, but I just find some bits and pieces here and there and I put it together into a product and that's what I love doing.

Demetrios:

Excellent.

Ulysse Maes:

But yeah, I don't even have a technical background. I studied business engineering in Leva, which is a small city in Belgium. I think 50% of the population is a student, actually. So very special vibe there. A lot of parties. I don't know. Does business engineering exist in the United States? No.

Demetrios:

I was going to ask what that is. We have financial engineering and things like that, but what is business engineering?

Ulysse Maes:

It's basically you're learning to become the bridge between the engineers in an organization and the upper management. Many of those business engineers aim for CEO positions in a later stage of their career. So you get a little bit of the technical stuff, you get a little bit of the financial stuff, and you try to combine it. Fascinating. But we had some programming courses, but it's very basic. It's not enough to do anything practically with it, but since a young child, really, I was always into building shitty, crappy HTML websites with iframes and stuff, with HTML two or something. While HTML five was already out I just didn't know, but I kept on doing that. And while I was studying, I started to do.

Ulysse Maes:

I think this is very similar story to many other solopreneurs or indie hackers. I started as a freelance software developer to earn a little bit extra on the site, and when I graduated, I felt like I wanted to do something for myself because I was always working for clients and I love doing that, but I had these ideas in my mind, and then I started some projects just for fun, really. I learned a lot of that. And the thing is, if you're building something for yourself, you can experiment a lot more, and that's great. Eventually, I lost focus a little bit for a while in the programming field because I started making music, and it really started to consume suddenly. I did it as a joke before, but then I was suddenly featured in a Spotify playlist. Oh, wow. Which was really cool for me at the time, during Corona, so I had nothing else to do.

Ulysse Maes:

I just started building a music career. I guess it wasn't meant to be eventually, but the thing was, scribewave is created because of this career, because I was making these songs, and I knew I had to promote them on social media, and I liked making music, but I didn't like recording myself. I didn't feel comfortable, so I started making lyric videos, and so I had content to put on TikTok, to put on Instagram reels and stuff. Problem was that it was a very tedious, boring job, and I thought, okay, I can build stuff, so why not build an automatic lyric generator?

Demetrios:

Nice.

Ulysse Maes:

And yeah, that's really the beginning of scribeweave. In the beginning, there was no speech to text automation. It was just me pressing the spacebar every time a word was said, which was really terrible, but was already better than doing it in premiere pro. So that was the first step. And then, because the gains weren't that much, I forgot about it until I discovered whisper. And then I thought, okay, I have to try this out. And it worked so well, I didn't even have to separate the audio tracks. I could just upload my entire song and it would recognize the words in Dutch because I sung in Dutch.

Demetrios:

Oh, wait, what's the story there? So you're in Belgium singing in Dutch.

Ulysse Maes:

Yeah. So in Belgium, there are three languages, and the biggest one is Dutch, actually.

Demetrios:

Really? Okay, so there's. Let me see if I am properly cultured. There's flemish, right?

Ulysse Maes:

Yeah, yeah.

Demetrios:

There's Dutch, and then there's English.

Ulysse Maes:

German.

Demetrios:

Oh, German, of course. How could I forget? I was thinking English because of the EU there. So I figured, oh, yeah, well, it's got to be English, right? Because everything happens there. Okay, so you were doing your thing in Dutch. You put it through whisper. You didn't have to do anything. It spit out the lyrics, and it said, is this right? And the accuracy was surprisingly high, I'm guessing.

Ulysse Maes:

Yeah, it was. I think there was. At that time, it was a word error rate of about 10%. But I was using whisper medium, my GPU, I don't even know if I had a GPU in that computer. I think I was running it on CPU, so it was really slow. But I was blown away by the results. And that was the turning point. That was a moment that I thought, okay, I need to build a product out of it.

Ulysse Maes:

And I bought a domain, songtovideo.com. I told it to a couple of my friends, but apparently it turned out I was the only one with this problem.

Demetrios:

So let's see, you have some artistic friends who also were thinking, oh, yeah, I want to make videos that are lyric videos. Because it is true that lyric videos got very popular, and it was almost like a being. I like to call myself a musician, too, although some people just call the music that I make noise. But that is a whole nother debate that we can talk about later. When you put out a song, you want many different ways of putting it out, because any of these different ways of putting it out can trigger the algorithm, right? So you want a full on video, like a traditional music video that they used to play on MTV, but then you also want a lyrics video that can be something that people can read along with. You also want maybe those short videos, so you want all these different options to play with, right? And that you realized you could do with whisper, and you could create a tool that would do that. And then you went out and you started asking around if anyone else had that problem, and you realized that they didn't.

Ulysse Maes:

Exactly. So that was a bummer at that time, because I really thought that I built something that all musicians would love. And then it turned out it wasn't that much of a problem for other people.

Demetrios:

And then you pivoted. You said, okay, I've got to pivot now because you got the bug in you, I guess. So you realize, well, I kind of like building things and seeing if there is traction. So did you pivot right to scribewave?

Ulysse Maes:

Yeah, I think that was the second domain I bought. It all started with my sister, because she was writing a thesis and she knew about my little project and she asked me if she could use the speech to text technology for transcribing her interviews because she had to do a lot of them for her thesis. And I was like, yeah, that should be possible. Was it completely off? It was not a use case that I had in mind initially, but I thought it was a nice challenge to build something for her and was also happy to help. I did this, I think I built a prototype in a day or two days, and she was really happy with it. I was blown away by her reaction. I didn't know it would mean so much to her, but she told me that she saved hours of time thanks to the little tool. Yeah.

Ulysse Maes:

Then I thought, okay, maybe I should develop this further and set up a payment link, see if other people think it's so much worth it that they want to pay for it.

Demetrios:

Then things got interesting. You had one happy customer and you thought, hey, maybe there's more of these out there. And did you totally shelf text to, what was it?

Ulysse Maes:

Songtotext AI songtovideo.com I guess I wanted to buy the AI domain, but I was too broke. It was amazing. I think it's $50 or $60 a year and the.com domain is only around ten. So the cheapest option.

Demetrios:

Your sister has a happy customer, you set up the payment link and then what happened?

Ulysse Maes:

Yeah, so now that I tell it, I can recall that before setting up the payment link, she already told me that she wanted to pay for it because at that moment I wasn't really thinking of turning it into a product or anything. I told her that, okay, I was running it that time. I was running it on serverless GPUs. So I uploaded whisper on banana, I guess. And yeah, that was working well at no scale at all. But yeah, I told her that it costed me a little bit of money and she was happy to pay for it. And that was really the pivotal moment where I thought, okay, I'll try if other people want to pay for it as well. Or is it maybe just because she's my sister?

Demetrios:

Could be.

Ulysse Maes:

It's like a little bit of a mom dust, I guess.

Demetrios:

One funny thing I want to point out there is that I wouldn't expect a university student, she's in university.

Ulysse Maes:

Yeah.

Demetrios:

They're not known for being flushed with cash, so I wouldn't expect a university student to offer to pay for something if it wasn't very valuable. And so I do see that it's, huh, maybe there's something here. If university students are saying that they'll pay for. Well, or maybe it's totally different in Belgium and people are working and getting money while they're working, and it's a different story than the prototype or the stereotype I have in my head.

Ulysse Maes:

No, you're right. That was for me as well. It was a good sign. If students want to pay for this, there are other people with a little bit more capacity to pay that definitely want to pay for this. So it was a good sign. Yeah.

Demetrios:

Excellent. And so you set up the link, you said, all right, we're open for business, and then the money just started rolling in.

Ulysse Maes:

Not exactly. It's still a challenge, though. But, yeah, I opened it up and it's a shame, and I can't really remember how it went, but I remember that three days later or something, the first payments came in and I didn't really know where it came from. Someone just found out, I guess.

Demetrios:

Oh, wow. And again, this was like, what, 2022? Late 2022?

Ulysse Maes:

No, it was early 2023.

Demetrios:

Okay, cool.

Ulysse Maes:

So a couple of months ago?

Demetrios:

Yeah, like eight months ago, basically, yeah.

Ulysse Maes:

So I contacted that person, I asked her why she bought credits and turned out she was a student as well.

Demetrios:

Friend of your sister?

Ulysse Maes:

I don't think so. But now I don't really know where she came from, but I probably posted it on my Instagram or something.

Demetrios:

Oh, nice.

Ulysse Maes:

It was a friend of a friend somehow. Yeah. But I can't remember exactly, which is a shame. I could look it up.

Demetrios:

But the key is that you said, I'm going to just start marketing this within my circles and see if there's any traction before I dedicate a whole lot of time and effort into it.

Ulysse Maes:

Exactly. So at that moment, it was a very rudimentary product. You just uploaded an MP3 file. I think it only supported MP3 at that moment. Yeah, I think MP3 it was. And then you got your transcript and that was everything. It was really basic, but people wanted to pay for it. So that was a great validation for me to put a little bit extra time in it and create a real product out of it.

Demetrios:

And talk to me about the product development and where you took it and who you were building for, because it seems like you had 100% of your users being students who were looking to do things in the academic field. Is that what made you want to focus in that area? And did you go hands down in that direction?

Ulysse Maes:

Not really, because like you already mentioned, students aren't. I just don't expect them to be very loyal customers. They only need it a couple of times, probably for their thesis and then they move on. So I had this hinge that I needed to focus on people that would need transcription on a more regular basis. So that's the second thing I did. I started reaching out on Twitter. I just looked for transcribe or transcription in Twitter and I checked some posts about it. And then I found a journalist who was complaining that no decent transcription solutions existed for the dutch language.

Ulysse Maes:

And I thought, okay, this is perfect guy to introduce my little tool. To turn out. He was really happy with it and he was really enthusiastic. He never became a customer, but he was really the person that made me believe that there was something in the media space. And I really wanted to show off a little bit. So I wanted to create a really good product to show him. So that was my motivation. And in about two weeks I rebuilt a song to video platform into a document based transcription platform for him.

Ulysse Maes:

Okay.

Demetrios:

And then the product evolution has been what until now. How did it go from that into what it is today?

Ulysse Maes:

I started adding more features for different clients. So there were content creators as well that came on later on. And then there were people that wanted to do translations, there were people that wanted to export to Google Docs, stuff like that. So I just added all this stuff. And sometimes I thought, okay, maybe this is a little bit too specific of a request. And I think sometimes we just need to say no if you're building such a product. I didn't at that time, I just built it, but I only enabled it.

Demetrios:

For that person on that account, no feature flagging it.

Ulysse Maes:

Yeah, but it was terrible. I never built something in production really. I only built stuff for myself. So at that time it was horrible. I just hard coded the IDs in the code and if this user with this ID is logged in, they show this part. So yeah, that was terrible. But eventually I opened up some other features that I just kept private to some users. I opened up for everyone.

Ulysse Maes:

I started tracking a little bit more what people were doing with it. And yeah, around one month ago I decided to start from scratch, but knowing everything I know now, and that really helped me to build a second version, which is so much better than the first one, I think was a good.

Demetrios:

Learning process because now you don't have all this technical debt of random hard coded features. For certain users, it's the worst. And now you didn't go out and build a team, you've stayed solo, right? How has that been?

Ulysse Maes:

I forgot to mention that I'm also doing a PhD. So I graduated and then I started doing a PhD. So you could still say it's a side project, although I'm working it on it on a full time basis, you could say. So in the beginning, I think it just grew so organically that I never really thought about starting a team. Now it's to say that in the last couple of months, I've started to look at some people, maybe to join a team also still looking for a co founder, but it's really hard to find someone that feels completely right. And it's also a me problem, I guess, because I don't have a set of requirements exactly for what I'm looking for in a co founder. I think it's, again, something that I'm learning as we speak. I'm constantly learning every day.

Ulysse Maes:

I learn so much by doing this. And this more personal, interpersonal aspect is also something that I don't have experience with, and I'm sure it will come along my way. So I've already met so many interesting people. I'll meet more people as well. And I'm certain one moment there will just be this click and I'll find the right person to join the team.

Demetrios:

And now, how have you been growing? Because I do appreciate that you've been growing significantly and as you say, it's a part time thing, but it could easily be a full time thing. I can imagine. And it's not hard to get the traction that you've had and then go and raise a big round of venture capital and say, no, I'm going to change my whole life direction, but take me back to like, are you posting on indie hacker forums? Are you building in public? What has been some of the biggest growth levers that you've found for me.

Ulysse Maes:

At this point, the biggest growth levers are really those AI indexing websites that suggest a certain tool for your use case. And apart from that, there's a ton of things I need to do and I'm not doing right now. So, for example, building in public, I really want to do that. I've got at this moment, four or five Twitter followers. So that's a bit of the issue, I think, to kickstart this every time I'm about to start it, I think, yeah. Will it make a difference? I've got five followers.

Demetrios:

One is your sister, I imagine, one.

Ulysse Maes:

Of which is my sister. Yeah.

Demetrios:

She's been the loyal fan from the beginning at.

Ulysse Maes:

So I think that's maybe also a bit because I'm based in Belgium right now, which is a very uninterpreneurial country in general, people are very risk averse. They don't like to stick out of the. How do you say that?

Demetrios:

Yeah, like, be the tall grass that gets.

Ulysse Maes:

So you don't want to be different. Basically. Of course, there are people that are super excited and super psyched about everything. That's technology and innovation, but it's really hard to find these people. On the other hand, it's such a small country that it's easy to reach whoever you want. So for me, the major part of my growth strategy up till now was just contacting the biggest media players in Belgium and asking them if they would be interested in using scribe. And up till now, most of them said yes. So that's really cool.

Ulysse Maes:

And I think that's the advantage of being in Belgium, that those people are within reach.

Demetrios:

Yeah. So it's really recognizing that you have cornered a niche that is basically what you can put a flag in the ground and say, this is mine, this niche right here, I can conquer this because there's not a lot of people that are going after this. And I have the ability to cold call people in different areas that I think would be great customers for this. And so you didn't let your five Twitter followers stop you. I really can admire that.

Ulysse Maes:

Yeah. But still, I need to work on this b to c outreach, especially because I don't think scribe wave is very well known. And right now, since I developed the entire new platform, and I'm much more confident about the scalability of the product at this moment than I was three months ago, for example, right now, the new challenge for me has become marketing and growing awareness.

Demetrios:

And it is, correct me if I'm wrong, you have focused historically on the market that is in Belgium. But it's not like you have to stay in that market.

Ulysse Maes:

No. Yet it was because of this reachability of those companies that I initially focused on that market. But of course, that's the great thing about whisper. It supports more than 90 languages, so you can easily board it to other countries, other markets. But again, that's a little bit the thing about a startup. It's a constant experiment. So I'm still figuring out who exactly is my target customer, but at this moment, most of them are journalists. Yeah.

Demetrios:

It seems like if you've got a great thing, just double down on that, double down on the outreach, and double down on the journalist, ICP, or ideal customer profile, and then go as hard as you can in that direction.

Ulysse Maes:

Yeah.

Demetrios:

And the other piece that I wanted to mention, you said how you're much more confident in the ability to scale now because of the refractoring that you did, or basically you destroyed the old app and you started from the ashes, the Phoenix grew out of it. Right.

Ulysse Maes:

That's a better.

Demetrios:

What are some things that you did to make it more scalable and make yourself feel more confident in the tech behind what you're doing?

Ulysse Maes:

First thing that really changed a lot was not running my own models anymore, but going to Deepgram, that was the biggest issue in scaling the product. Initially I used serverless GPUs for running the model, but there was always a cold boot time. Half of the time it wasn't available. I think it was also because at that point, also the platform that I was using was very early stage, so might be a little bit more stable at this moment, but still, it was something that consumed a lot of my time. So I'm very happy to outsource that part so I can focus on building the features that take this transcript as a given and help you as a journalist, for example, or a researcher, to do stuff with this transcript. That was the first big scalability improvement. And then second thing is that I optimized my AWS environment completely. When I started this project, I didn't know what a lambda function was.

Ulysse Maes:

So I learned a lot there, which.

Demetrios:

Is what for those of us that.

Ulysse Maes:

Do not know what it is, a lambda function is a serverless function that you run on AWS, so it can infinitely scale, which is really nice when.

Demetrios:

I scale up and scale down. Right?

Ulysse Maes:

Yeah, that's also important. So you're never using too much compute, but you're always able to serve your customer. I first did it by just running back end on my computer, on my laptop, and I let it on, I played it in the socket of the university, was connected to the network, and then I thought, okay, I've got a server now. Wasn't a great idea.

Demetrios:

You found that one out fast, I imagine.

Ulysse Maes:

Yeah. So, yeah, those are the two biggest scalability improvements. And then apart from that, I used the framework mentine for my front end, which also enables me to create a much more consistent layout. In the beginning, everything was all over the place, so everything was styled in CSS, but it was just custom CSS. I didn't have a global style or anything. I just thought, okay, 25 pixel spacing would be nice here, maybe 43 here. So in the end it didn't really add it up. So it's better to use a framework.

Ulysse Maes:

And it's a tip that I could give to everyone doing this, use that.

Demetrios:

From the start, don't hard code features into the different users.

Ulysse Maes:

Yeah, I think if there is a list of the seven worst practices that you could do, I did all of them the hard way, which is a.

Demetrios:

Testament to, I guess, how good the product is considering you were able to do everything, in your words, wrong, but you still are able to get more people using the product and get people interested in what you're doing and keep growing. So I love that you don't let that stop you and you learn from what you can. Then you say, okay, now, next time I do it, I'm probably not going to do it like that. I'm going to make sure that I optimize it a little bit more or I perfect, I iterate and I get better. So this has been incredible to talk with you, man. I really appreciate your viewpoint and again, your dedication. It is really cool to see the drive that you have and what you've been doing with scribewave. I encourage everyone to check it out at least and hopefully give you a follow on Twitter so that you have maybe seven followers after this one.

Demetrios:

Yeah. If anyone wants to reach out to you and get to know you more, we'll leave all the links in the description so that they can contact you. And maybe, who knows, you might have your future co founder in the audience listening to us.

Ulysse Maes:

Perfect. Yeah. Thank you for having me. Was super nice to have this talk.

Demetrios:

All right, I'll talk to you later. You.