How we built and launched the AI audio note app SpeakMemo in 5 days, and what we learned.

Hello everyone, I'm Armin!

Recently, my friend Beck and I have been working on SpeakMemo (https://speakmemo.co/), a Telegram bot designed to record and automatically transcribe your ideas into bullet point notes.

This post aims to share our experiences during the development process, and hopefully, it will provide some useful insights for you!

Why We Made SpeakMemo

Daily, I find myself taking a lot of voice notes. Whether I'm meeting new people and recording our conversations, or capturing product inspirations during a walk, I tend to transform my experiences into voice notes.

However, I struggled to find a product that could efficiently transcribe these voice notes, a tool that could quickly capture my spoken words and reformat them into text suitable for my note-taking app like Notion.

When I discussed this problem with my friend Beck, who also frequently records his daily activities, he suggested, "Why don't we build one ourselves?" It was a compelling question, and thus, SpeakMemo was conceived.

How We Built SpeakMemo Within 5 Days

Once we committed to developing this product, Beck and I decided to swiftly construct a Minimum Viable Product (MVP) to introduce it to the market and gather user feedback.

Our MVP was designed with the following features:

• Users can record audio.
• The audio is transcribed into text, which an AI then reformats into bullet points suitable for note-taking.
• Users can then copy the transcribed text and paste it into their preferred note-taking app.

For quick launch, we leveraged various third-party services to help us construct the product instead of starting from scratch:

• Telegram Bot: We started with a Telegram bot due to the ease of development. With just a few commands, the bot was operational, and it already came with recording functionalities.
• Whisper: We utilized OpenAI's Whisper technology for transcribing the user's voice, which worked excellently and supported multiple languages.
• ChatGPT: We employed ChatGPT to reformat the transcribed voice notes. It eliminated pauses, intonations, and repetitive words, converting them into clear, concise text.
• Backend & Hosting: After developing our Node.js application, we required a platform that was user-friendly, provided a great developer experience, and had robust built-in features for deploying applications. This led us to choose Railway.
• Supabase: We used Supabase for data storage. As our database, Supabase is a powerful and flexible platform that enables developers to quickly and easily build serverless applications.
• Framer: We used Framer to create our landing page. Similar to Figma, it allows easy web page design with a drag-and-drop interface. We found it simpler to use compared to Webflow (and a fun fact, Framer recently launched AI-generated landing pages, you might want to give it a try!).
• Stripe: We integrated Stripe for handling payments. As a subscription-based product, we also offer a limited-time Lifetime Deal for our early supporters.

Lessons Learned

• Solve your own problem: Both Beck and I are avid voice note users, and the need for a product like SpeakMemo arose from our own experiences. When we shared SpeakMemo with our friends, we discovered many of them found it very useful!
• Ship fast: We weren't keen on spending six months building a product without validating user needs. Fortunately, numerous third-party services allowed us to develop our product in just five days! If the product fails, you've only invested five days, not half a year.
• Keep it simple: The initial version of SpeakMemo was just that - simple! While we have numerous ideas for a web version, a mobile version, and additional features, implementing them all would require significantly more time. The first version of your product should be straightforward, allowing for rapid iteration based on user feedback.

I am very willing to share our experiences on building https://speakmemo.co/ with everyone. If you have any questions, feel free to reach out to me at https://twitter.com/armin1i