How I Left My Job To Build An AI Podcast Editing Tool

Adrian Spataru

•

Published: August 13th, 2022

save

Adrian Spataru

Founder, Cleanvoice AI

$400

revenue/mo

Founders

Employees

Cleanvoice AI

cleanvoice.ai

from Bucharest, Romania

started May 2022

$400

revenue/mo

Founders

Employees

Discover what tools Adrian recommends to grow your business!

Discover what books Adrian recommends to grow your business!

Company of One: Why Staying Small Is the Next Big Thing f... Paul Jarvis

Zero To Sold: How To Start, Run, And Sell A Bootstrapped ...

📗

Want more updates on Cleanvoice AI? Check out these stories:

Hello! Who are you and what business did you start?

Hi! My name is Adrian. I studied Computer Science and I used to be that friend who you would call to fix your PC for free. However, I wanted to be a cool guy, so I start learning machine learning and AI.

My interest in applied AI, motivated me (after a lot of projects) to start Cleanvoice, a tool that automatically edits your podcasts with the power of machine learning. It removes filler words, stuttering, and more from your audio. In essence, it edits the stuff which you normally cut out from your podcast.

It's designed for podcasters and editors, who just want good edits without spending hours doing it themselves.

Pitch Talk from my previous project

Currently, Cleanvoice makes about 400$ MRR. We have a SaaS business model but also a pay-as-you-go credit-based pricing, which allows customers to use the product, without a subscription hell.

What's your backstory and how did you come up with the idea?

Like every millennial, starting a podcast in 2020 was the thing. During the holidays of December 2020, I started a podcast with a good friend of mine. Both of us were working as Data Scientists at the time, so we talked about uncommon AI applications (such as AI in Anime, Agriculture, etc). Recording the first episode took less than an hour, however, editing took over 5 hours. Since both of us were not native English speakers and we both stumble our words often, editing really took a toll on me.

Yes, weird is written incorrectly

I started searching for podcast editing tools, more specifically a tool which would remove all the "uhhh" and "umms". I tried tools such as Descript, but the issue is that if you don't have a perfect RP British or Standard American accent, it just didn't work. Since my co-host had a strong Ukrainian accent, there was no tool that could save me.

After 3 episodes, I got sick of editing and I decided to build an AI which would detect our “umm's” and “uhh's” in our conversation.

Cleanvoice was not my first business idea. I already invested a lot of time working on an AI Icon Generator tool called bezier.ai. I got contacted by Adobe for an acquire-hire opportunity, however, it wasn't successful. Emotionally drained, I felt a need to take a break. Cleanvoice was a great opportunity for a restart.

embed:tweet

After 2-3 weeks, I had a basic prototype and reduced the podcasting editing time down to 30 min. I tested the algorithm with other voices and sometimes it worked well. After converting my first customers, I abandoned bezier.ai completely and focused on my next business venture on the side. After a year, I decided to leave my stable job and focus fully on Cleanvoice.

Take us through the process of designing, prototyping, and manufacturing your first product.

Since we already recorded 3 episodes, I had roughly 3 hours of audio material in which I could train an AI. I started to dedicate one week to research, spending every day reading research papers on Audio Processing. After figuring out how I would build the main Architecture of the AI, I started coding and training the algorithm for a week.

To my surprise, it kinda worked, but had many flaws. The biggest surprise though was that it worked also with other voices. However, this was not enough for an MVP. AI was the easy part, but building a web app, billing system, and all the bells and whistles you see in apps these days, was unknown to me.

I didn't want to learn any complex frameworks that “professional” web developers use, so I made something scrappy that was stored in a single .html file.

For billing, I used Gumroad, since it was easy to integrate billing with just a single line of code. The pricing, back then, was usage-based.

Basically, you could upload any audio (as long as it was under 100 MB) for a single credit. The initial prototype was focused only on removing uhh and umms. Since then other features such as stutter removal, dead air and more have been introduced.

Since AI is resource hungry, I rented a cloud server for 40$ per month in order to make the MVP run without crashing.

Describe the process of launching the business.

After building an MVP, I didn't want to launch it for free. But I also wanted feedback from as many people as well. The solution was to make a paid app but gave away free keys for anyone who asked.

I went to various podcast-related Facebook groups and posted the following:

These posts were getting a lot of attention, and all I did was just send license keys through Facebook DMs all day long. Interestingly, some people also purchased keys, even though they could get one for free. I have received a lot of feedback and overall it showed me that there was a need for such a product.

In the first week, I made $150 in key sales. Afterward, I stopped advertising, and I tried to make the business legal, this was tricky since everything was closed down due to the pandemic, and it took me about 3 months to actually incorporate the company.

During this time, I stopped charging money and offered a “free week” campaign where people could use the tool for free. The week became a few months until I managed to set up my billing system and started charging again. Furthermore, I stopped advertising it, so I could focus on building a more usable product. Still, some users were still using Cleanvoice and when I started charging again, the majority of people converted to my first paying customers.

Looking back, this was overall just silly and stupid. I should've kept the momentum, but I didn't know better back then.

Being 24/7 for your clients is difficult, but it’s the best way to do customer discovery.

Since launch, what has worked to attract and retain customers?

Since I don't have much data, it is hard to tell exactly. But, my main acquisition channel, for now, is my podcast name generator. It ranks on the first page of Google and is my main source of traffic for Cleanvoice.

SEO was something that I haven't done for almost a year. (Though my excuse was that I didn't have enough time, that's what I lie to myself )

Writing content is not my forte, but creating AI tools is. So I searched for keywords which I could build tools for. Building free tools is my main source of traffic, and they bring most of the signups. Keywords with the word “generator”, “free app” etc, tend to be high volume with relatively low difficulty to rank.

Another thing was going to various podcasts as a guest. I’ve created an account for matchmaker.fm and contacted podcasts related to podcasting or entrepreneurship. Writing a personal message to each podcaster, and checking their previous episodes, really helped to understand what the podcast host is searching for. People generally spam on matchmaker, so this definitely will make you stick out.

Hacker news Launch

I checked the top-ranking Show HN posts on hackers' news and observed a pattern.

First is that the titles were self-explanatory. You knew what the tool did, without even seeing the landing page.

Second, there was always comment in the post by the author, where they explain the backstory of the tool.

I have no idea if this is the formula for success on hacker news, but at least for me, I landed on the first page on hacker news. This brought a lot of traffic to the website, but to my surprise, after a week. Only 2 people converted.

Start with SEO before you even launch. It takes time, so why not start now?

To no one's surprise, not many tech people host podcasts. Since then, I am more focused on the quality rather than the number of users.

How are you doing today and what does the future look like?

If you consider only the operations costs (and not my time), then it's currently breakeven. However there was a lot of time invested into it, so it might take a while to see those gains back.

Marketing is definitely something that I have ignored in the past. Being an engineer who likes to focus on the product more than on marketing. A trap that I fell into, like many other developers.

My focus now is on building YouTube videos and on the things which currently working. In order to avoid investing too much time into products, I split my weeks into marketing and development weeks. Where one week I focus on marketing activities and the other on my AI research. The more time I invest in marketing, the more returns I get (duh.) But as shown above, SEO is improving, and hope my new marketing venture will do the same.

Through starting the business, have you learned anything particularly helpful or advantageous?

SEO! But I had 0 knowledge of that a year ago. The first rookie mistake was building a single-page website. This was bad since I was barely ranking for anything except for the word “cleanvoice”. Creating dedicated pages for every feature/use case of my product improved it significantly.

Starting a business is hard. And when you are in SaaS and have clients from various countries in the world, support is something that can take your focus away. However, by allowing my customers to reach me very easily, (If you write me in support, you'll write directly to my phone) I learned a lot about their needs. Yes, being 24/7 for your clients is difficult. But I don't regret it. It’s the best way to do customer discovery and better than trying to schedule calls.

What platform/tools do you use for your business?

Plausible.io - Privacy-friendly web Analytics. No need to ask for cookie permissions and also an easy-to-read dashboard.
Crisp - Customer Support. Super convenient when you are not home and do text support on your phone.
Paddle – Merchant of Record. I live in the EU and dealing with VAT is something I don't want to do. Paddle just sends me only 2 invoices at the end of the month, which makes accounting super easy.
Kanuu.io – building a Billing UI is not hard, but for me, this is too much effort. Kanuu offers a plug-and-play solution for Paddle, and I have no complaints.

If it weren’t for Paddle & Kanuu, I would potentially never launch.

Vercel - Hosting the Website. Super easy to deploy my website. Several clicks and my app was live.

What have been the most influential books, podcasts, or other resources?

Company of One - Business book focusing on solopreneurs.
Zero to Sold - Certain chapters helped me get over the decision fatigue.
The Mom Test - Interviewing Customers.
Can't Hurt Me by David Goggins- Helped me to stay motivated in business and in my pushups.
All the content from Basecamp/Signal37. Their approach to work inspired me to start a startup.

Advice for other entrepreneurs who want to get started or are just starting out?

I am not qualified to give anyone advice. But maybe learn from my mistake and start with SEO before you even launch. It takes time either way, so why not start now?

Also, get a whiteboard! Whiteboard has been my go-to tool for brainstorming. As my girlfriend says, don't waste paper on something that won't work.

Where can we go to learn more?

Website
Twitter
Want to follow my entrepreneurial journey? Listen to my new podcast

If you have any questions or comments, drop a comment below!