I Created A Niche Text To Speech Tool & Hit 80K Users
I’m Gojko Adzic, author of Running Serverless, Impact Mapping, Humans vs Computers, and a few more books. My book Specification by Example won the Jolt Award for the best book of 2012. I’m one of the 2019 AWS Serverless Heroes, the winner of the 2016 European Software Testing Outstanding Achievement Award, and the 2011 Most Influential Agile Testing Professional Award. I’m the founder of Naraket and one of the co-founders of MindMup.
Narakeet is an online service that helps people make narrated videos easily. It is a text-to-speech video maker, aimed at people who are not video professionals, such as developers, online educators, and marketers. In brief, it turns PowerPoint presentations and markdown files into videos, creating life-like narration from speaker notes and voice-over scripts. Users can edit video as simply as they would edit text, saving hours by not having to record and re-record audio, synchronize pictures with sound and transcribe subtitles.
The product was launched in public beta in April 2020, and commercially in October 2020. Roughly 80K users created 220,000 videos since the launch. Narakeet now supports 40+ languages and local accents, with more than 250 high-quality neural engine voices. Neural-network AI voices are so good now that they can sound better than people who are not native speakers of the target language, or even native speakers with a strong accent. Text-to-speech narration is also amazing for people who need to record a voice-over but do not have professional audio equipment or a sound-proof environment, or are not pro voiceover artists. Narakeet helps to make videos faster by automating most of the voiceover and editing.
What's your backstory and how did you get into entrepreneurship?
I’ve been fascinated with programming since early childhood. Making products with software is the closest thing to magic today. Like wizards from folk tales, we turn some cryptic words into stuff that people can interact with in real life.
Plant the seed somewhere, then see how it takes root, then water it with lots of attention and user feedback. Growing products like that are very satisfying.
I started by copying PEEK/POKE instructions from Commodore 64 magazines as a kid in the eighties, making the computer wake up my folks with weird sounds. In high school, I competed in demoscene assembly programming, and then halfway through university realized that I can make money doing stuff I loved. I went through a bunch of programming jobs in the late nineties and early 00s then started working for myself. I created a consultancy business that helps large organizations, mostly financial institutions, improve software delivery (It’s still going strong, although these days my partners have taken over most client engagements.) Somewhere along that journey I ended up realizing that the kinds of jobs I was working were not fun anymore. The money was great, but the fun was gone. Then I started building my products and rediscovered the joy of programming.
In 2013 I co-founded MindMup, an app that millions of students and schoolchildren use to capture ideas, structure writing assignments, and organize thoughts. In 2020, I launched Narakeet. In the meantime, I also co-founded and built a bunch of stuff that didn’t do that well, including a social games platform, a post-conference engagement app for event organizers, a developer-oriented testing tool, and a video game about what would happen if East Germany survived till today and Stasi ran Google and Facebook. Most of my businesses were bootstrapped, funded initially by my consulting work or other stuff I built.
Take us through your entrepreneurial journey. How did you go from day 1 to today?
In addition to all this other stuff, I also like making open-source products. One was Claudia.js, an open-source tool for deploying Lambda functions to AWS. Claudia was one of the early tools that tried to make the process straightforward while AWS was still getting its act together around serverless. The key thing there was developer experience, and I wanted to make video tutorials for Claudia so that people could get started easily.
Coming up with the video content was easy and fun. The rest of the experience was frustrating. Aligning pictures and synchronizing them with sound felt like a huge waste of time. Recording narration was even worse. It took me seven attempts to do a single paragraph without mistakes. It still sounded bad, so I gave up and paid a professional voice artist. Her voice was great, but she spoke slower than me, so I had to re-edit the videos again. For one video, it took me more than four hours of effort to create just five minutes of good content. Six months later, a web page is shown in that tutorial significantly changed, and I wanted to update it. Instead of a simple tweak, it all came crashing down. I foolishly tried to save money by asking the artist to record just the changed part. It was the same person, the same equipment, but you could hear the difference. I had to choose between a video that sounded inconsistent, or paying for the whole thing again and wasting time on re-editing everything. There had to be a better way of making videos, but I couldn’t find any, so I decided to build it.
It all started with a bunch of shell scripts, and soon it was taking me ten to fifteen minutes to make a great five-minute video, not hours like before. Around the time the video workflow came together, YCombinator decided to open up its Startup School program to a remote audience. I wanted to participate but needed a product to take from the idea to launch in a few months. I thought that other people who are not video professionals must have the same problems I had and that it would be fun to try turning my shell scripts into a product so others can benefit as well. That’s how Narakeet came to be.
Narakeet takes care of all the boring and time-consuming tasks of video editing, and lets users focus on creating good content. It will automatically create video clips from images, align audio and video segments, add captions, display text on top of videos and apply transitions between scenes. It will record life-like narration using neural text-to-speech systems, so the audio will be consistent no matter how many times you change the script.
The tool was originally intended for techies like me, who like to script stuff. Having videos built from text, with support for version control, continuous integration, and building pipelines is very liberating for people like that (it even supports GitHub actions so you can rebuild videos automatically when your screenshots change).
The tool found its way to a much wider audience that has the same problem, but can’t use any markdown, GitHub, or even cares about continuous integration. This crazy covid pandemic made video lessons more important than ever before, and teachers all of the sudden had to become good at video editing. They rarely have the skills or the equipment to make good videos, let alone the time. (One NCTQ study found that teachers can afford to spend, on average, about 45 minutes a day planning lessons.) A bunch of teaching sites started recommending my tool and I discovered that there’s a much wider market for it than I originally thought.
After a lot of user research, an idea emerged to just let people convert slides into videos. Teachers already prepared lectures in Powerpoint or something similar, and they’d usually type up the stuff they wanted to speak about in presenter notes. Narakeet just took that idea and did the next logical step. It turns slides into a video by using the presenter's notes for voice-over. I knew that we had a good product after receiving an email from one of the university professors who participated in the public beta. The professor was amazingly enthusiastic about the whole experience, saying that he created a video lecture in under three minutes, and estimating that just recording and editing the audio would have taken him at least three hours. Another beta participant was raving about how useful the tool was since her neighbor had a loud dog who’d bark all the time, and that would come out in the audio she usually recorded. Using text-to-speech narration helped her create a much better voice-over, significantly easier.
How are you doing today and what does the future look like?
Narakeet is profitable, growing mostly through word-of-mouth and happy customers recommending it to others.
The product operates on a freemium model. To reduce the barrier to entry, website visitors can create videos with restricted capacity without even registering. Users can pay for larger capacity and longer videos. The conversion rate from free to paying customers is around 2% today.
Like most of the other stuff I’ve launched, Narakeet is bootstrapped so it was originally designed for the very low cost of operation, and a quick route to profitability.
The plan for the next year or so is to expand the number of available languages and voices significantly, so it becomes the biggest library of online text-to-speech voices.
Through starting the business, have you learned anything particularly helpful or advantageous?
My biggest lesson with Narakeet is how important it is to sort out the trademarks early. Because it started as a way to scratch my itch, then turned into a side-gig, it took a long time (almost a year and a half) for the product to reach a point where it started to look like there is commercial business potential. At the point where it became clear that it has a commercial future, I applied for a trademark on the name, and this was far too late.
The product originally had a different name, and the lawyers for a large company objected to our trademark application, claiming it was similar to theirs. The advice I got at the time was that there’s a good chance I could win the dispute, but that the process could drag on for months or years and potentially be very expensive for me. The objection came from lawyers representing a multinational giant with a big war chest, on the very last day they could object,
so they were playing the game of wasting my time. The start of the new school season was just coming up, and I wanted to launch the product commercially in time to catch that window of opportunity. I figured that I have a lot more to lose by delaying, and decided to back down and rebrand.
By that point, the current product name was already out there for almost two years, and the rebrand ended up costing a lot in terms of search performance and user visibility. Although setting up a redirect is easy, it looks as if Google penalizes brand-new domains, so the product visibility from search plummeted. Even today, people are still searching Google for the product under the old name, reading about it from old blog posts, and a good portion of them don't click through to the new website (only about 40% of people do, according to our Google Search console).
It’s difficult to put an exact figure on this cost, but it had a huge negative impact on growth exactly at a time when the product needed a boost. That’s pretty easy to see from the website visitor statistics.
Next time, I’ll get a trademark as soon as something becomes a viable idea, not waiting until it’s ready to launch commercially.
What platform/tools do you use for your business?
Narakeet runs on AWS serverless infrastructure (Lambda). It’s amazing because it is autoscaling and the costs are directly proportional to usage. For any kind of business expecting growth, it’s key to have enough capacity to service spikes in demand. On the other hand, as a bootstrapped business, avoiding static overhead or fixed operating costs is very important. Lambda provides a great way to meet both these needs.
There was a relatively popular education channel on YouTube that recommended Narakeet, and more than ten thousand people came to try it in a short period. I didn’t even know about it until I woke up the next morning, and everything worked perfectly. AWS scaled our stuff up to meet the demand, then scaled it back down after the visitors stopped making so many videos. It just worked.
Because billing is proportional to usage, it was just important to make the right decision about the pricing model so it covers the costs with a good enough profit margin, and the rest works out of the box.
What have been the most influential books, podcasts, or other resources?
YCombinator Startup School was highly influential for my work on Narakeet. I originally started the product as an excuse to participate in that program. The videos and resources from past years are available online, and they are a wonderful way to learn about lots of different aspects of running a software service company. I learned a ton about customer engagement from that.
More recently, the MicroConf Slack workspace turned out to be a great resource, being able to chat with other founders, get feedback on ideas, and find partners and service providers that can help. I solved lots of problems along the way by applying the advice from chats, often someone recommending something to another person that I didn’t even know I needed. Lurking in those virtual conversation hallways is amazing.
Jeff Patton’s User Story Mapping was critical to clarify the whole user journey and experience around Powerpoint to Video conversions. Jeff wrote a book about it that should be mandatory reading material for any software company founder and product manager.
Another book I’d recommend is Lean Analytics by Alistair Croll and Benjamin Yoskovitz - I love their Stages of the Growth model. It helped me clarify what kind of metrics to track and objectives to set at which point of the product life cycle. After reading the book, I realized that with some of my previous projects I set the wrong types of targets early on, damaging growth and adoption. Narakeet benefited significantly from clarity around that stuff.
Advice for other entrepreneurs who want to get started or are just starting out?
I’ve found it wonderfully useful to start by scratching my itch, but then listen to the community and let the product evolve in its direction.
With Narakeet, the original use case and the intended audience were significantly different from where it ended up. The product found a user base I never really expected, which had the same problem but needed a completely different way of interacting with the product. The original markdown to video use case is still supported, but more than 90% of the videos on the site are produced from slides.
Plant the seed somewhere, then see how it takes root, then water it with lots of attention and user feedback. Growing products like that are very satisfying.
Where can we go to learn more?
For more info, or to keep up to date use RSS, Slack, Twitter, YouTube, Facebook, Instagram or TikTok.
Download the report and join our email newsletter packed with business ideas and money-making opportunities, backed by real-life case studies.
Download the report and join our email newsletter packed with business ideas and money-making opportunities, backed by real-life case studies.
Download the report and join our email newsletter packed with business ideas and money-making opportunities, backed by real-life case studies.
Download the report and join our email newsletter packed with business ideas and money-making opportunities, backed by real-life case studies.
Download the report and join our email newsletter packed with business ideas and money-making opportunities, backed by real-life case studies.
Download the report and join our email newsletter packed with business ideas and money-making opportunities, backed by real-life case studies.
Download the report and join our email newsletter packed with business ideas and money-making opportunities, backed by real-life case studies.
Download the report and join our email newsletter packed with business ideas and money-making opportunities, backed by real-life case studies.
