Louis Pereira Built a Business That Turns Rambling Speech into Writing
How Louis Pereira built AudioPen out of a small AI experiment while also helping run the family business. The story examines the preparation that came before the 12-hour build, the early audience and paid demand, revenue from term-based passes, and the ongoing operating burden.
AudioPen is an AI tool that transcribes what a user says in a rambling way and rewrites it into sentences that are easy to read. It focuses on producing drafts of emails, notes, and writing. In an Indie Hackers interview published on February 17, 2026, founder Louis Pereira said his monthly revenue was $15,000 to $20,000.
A Small Experiment Started Alongside the Family Business
Pereira built internet products while helping with his family's offline business in Goa, India. The family business was a retail operation running supermarkets, hardware stores, and sporting goods stores, and he worked there during the day and built digital products at night. It was a case of growing a separate source of income while keeping the existing livelihood.
He did not come from a professional development background. By his own account, he had wanted to try an internet business since 2015, but he repeatedly started learning to code and stopped. The turning point was learning Bubble, a no-code development tool, in 2021, and he went on to build several products and gain experience turning ideas into working services.
AudioPen started as a small technical experiment. While learning how to use AI APIs, he put several tools on his personal website, and among them the feature that transcribed speech and cleaned it up drew a particularly strong response. Seeing how users reacted to the experiment he had published, he developed it into a standalone product.
The trigger for making it a standalone product was the Half Day Build event he hosted himself. It was an event where participants took an idea to actual revenue within 12 hours, and he chose AudioPen as the subject for that day. However, the well-known “12-hour build” does not include the earlier test feature or the screen designs prepared in Figma before the event. To judge the creation time of the business, that preparation has to be considered too.
What he wanted to validate was the minimum product useful enough that people would pay for it. Some of the people he asked to try it before the public launch bought the paid version, confirming real willingness to pay. He explained that the first product was built in about 12 hours, but it was then refined continuously over several months.
Where the First Customers Came From
The online activity he had built up beforehand contributed to early sales. He had been publishing his product-building process since early 2021, and in a 2024 retrospective he said it took about two days to win his first 100 paying customers. He cited consistent public activity and more than 10,000 X followers as the backdrop. What matters is not only the fast development but that he already had people to show the product to.
For the next wave of spread, a mention from someone else helped. According to an early interview, sales increased sharply after an account with around 200,000 followers posted about the product. On May 4, 2023, it also took first place for the day on Product Hunt, gaining additional exposure.
A Solveo interview covering the early results reported $73,000 in revenue and more than 1,000 paying users over the first two months after launch. The same interview put total signups at more than 11,000. Here the $73,000 is two months of cumulative revenue, not monthly revenue or net profit. Total signups and paying customers also have to be kept separate.
He maintained the relationship with existing customers through email. Pereira put personal explanations into the emails announcing product updates, trying to make users feel they were part of how the product changed. He also ran an affiliate program to encourage referral sales, and separately explained that spontaneous word of mouth from customers was a main channel of new arrivals.
Paid promotion did not produce the same results. In an October 2024 interview he said he spent about $7,000 on TikTok influencer marketing, but there was no increase in visitors during that period. Organic referrals early on and paid promotion performed differently, and at the time he was focusing on improving the product rather than expanding marketing.
What Users Paid For Was the Cleaned-Up Draft
AudioPen provides both a transcript that preserves what was said and a polished version of it. Users can choose the style and how much editing they want, and they can also add examples of their own earlier writing to get a similar style. It reduces the burden of having to compose sentences while speaking, while leaving room for users to review and fix the output.
Actual use ranged wider than simple voice notes. Examples the founder published in March 2026 include a software engineer drafting a complicated email by speaking, an executive capturing memories right after a meeting, and a user who speaks in Czech to produce a letter in English. The purposes differ, but they share the same core: getting a thought out in speech and turning it into writing another person can read.
The core conversion feature is available in the free version as well. Based on the official guidance checked on October 2, 2026, the free version supports up to 3 minutes per recording, and the paid Prime plan supports up to 15 minutes. Prime adds unlimited note storage, audio file upload, rewriting existing notes, and merging several notes. The core effect is shown for free, with paid features offered to users who want to use it more often or accumulate material.
He kept using no-code tools even after the product grew. The setup described in a February 2026 interview was Bubble for the web app, Draftbit for the mobile app, and Xano for the shared backend. The current official product is offered as a web app, mobile app, desktop app, and Chrome extension, and a single Prime account can be used across multiple devices. Starting from an early web tool, it widened its reach into the environments where users actually write.
Telling Term-Based Payments Apart from Monthly Revenue
The payment model was closer to selling long-term passes. In an October 2024 interview he explained that he charged $99 per year but did not use automatic recurring billing, and in a February 2026 interview he said only 1-year and 2-year options were offered. The official prices checked on October 2, 2026 are $33 for 3 months, $99 for 1 year, and $159 for 2 years, so shorter-term options were added later. The past interview's statement that only annual and 2-year products were sold should not be applied as is to the current plans.
Looking at the disclosed revenue by point in time, it was about $15,000 per month in the May 2024 Starter Story interview, $12,000 to $15,000 per month in October of the same year, and $15,000 to $20,000 per month in February 2026. These are self-reported figures from different points in time, not a continuous set of financial records showing growth month by month. In particular, there is no explanation of how the long-term fees were allocated by month, so the interview's “monthly revenue” should not be rewritten as strict monthly recurring revenue (MRR) or used to estimate net profit.
The cost structure was based on a small organization and the use of external AI services. The 2024 Starter Story interview described it as one founder and zero employees, and Pereira said it was profitable from the first day of launch and that the biggest cost was the AI API fees. However, no detailed cost breakdown or profit margin was disclosed, so there is no basis to conclude that most of the revenue became personal income.
Behind the Fast Launch Came Ongoing Operations
Along the way he also ran into attacks on the service. Pereira explained that a DDoS attack temporarily took the site down, and that some email providers even classified the site as spam. Given the considerable effort required to respond to users one by one and restore normal operation, a product that can be built quickly and a business that runs itself are two different things.
What stands out about AudioPen is the sequence: publish a small experiment, focus on the feature that got a strong response, and then confirm demand through actual payments. In his 2026 official retrospective, Pereira still described himself as a part-time indie hacker and wrote that he would keep improving the product. It is closer to a business that steadily ran a small paid product on the strength of long-accumulated building experience and an early audience.