Transcriber App

Mabel Anto Moncy — Sun, 19 Apr 2026 09:34:01 +0000

Audio Transcriber: Building Technology with Purpose

A Project That Became Personal

In a developer’s journey, there are countless projects—some built to learn, some to experiment, and some to showcase skills. But every once in a while, a project becomes something more than just code.

My Audio Transcriber App is one of those.

I built this application for my father, who is hearing impaired. Watching him struggle to understand voice notes and audio messages—especially on platforms like WhatsApp where regional language transcription is often unavailable—made me realize a simple but powerful truth:

Technology should be accessible to everyone.

This project was my attempt to bridge that gap.

The Problem

Audio communication has become a major part of our daily lives. From voice notes to recorded messages, people increasingly rely on spoken content.

But for someone with hearing difficulties, this creates a barrier.

Voice notes are inaccessible
Regional language support is limited
Important information gets lost

This wasn’t just a technical problem—it was a human one.

The Solution

The idea was straightforward:
Convert audio into readable text in a simple, usable way.

But to make it truly useful, the app needed to be:

Reliable
Easy to use
Accessible across devices
Capable of handling real-world scenarios

Key Features

Core Functionality

Real-time Audio Recording – Record audio directly within the app
File Upload Support – Import audio files, including WhatsApp .opus formats
AI-Powered Transcription – Built using Google Gemini 2.5 for accurate results
History Management – Store and revisit previous transcriptions
Smart Garbage Detection – Filters out empty or accidental recordings
Bring Your Own Key (BYOK) – Flexible API usage model
Dark/Light Mode – Clean UI with Material Design 3

Engineering the System

Building this app wasn’t just about adding features—it was about designing a system that works reliably under real conditions.

Triple Fallback AI Architecture

To ensure consistent transcription quality, I implemented a multi-layer fallback system:

Primary: Gemini 2.5 Pro (high accuracy)
Secondary: Gemini 2.5 Flash (optimized speed)
Tertiary: Gemini Flash Lite (lightweight backup)

This ensures that even if one service fails or slows down, the system continues to function.

Handling Real-World Failures

The app includes:

Network timeout handling
Graceful error recovery
Connectivity checks

Because in real-world applications, failure handling is just as important as success cases.

Why Serverless?

One of the biggest decisions in this project was choosing a serverless architecture.

The reason was simple:
I didn’t have the money and resources to maintain a dedicated server.

So instead:

API calls are handled directly from the client
No always-running backend is required

This approach has trade-offs:

Lower cost
Simpler deployment
Reduced security (client-side API exposure)

It’s not perfect—but it’s practical.

And that’s an important lesson:

Good engineering is not always about perfect solutions, but about making the best decisions within constraints.

Beyond Code

This project is more than just an app.

It represents:

Accessibility through technology
Solving real-life problems
Building with empathy

Every feature was designed with a purpose.
Every decision was influenced by a real need.

What I Learned

Designing AI-powered applications for real users
Balancing security, cost, and scalability
Importance of fallback systems in production environments
Thinking beyond code to create meaningful impact

Final Thoughts

Not every project you build will stand out.
Some will just be stepping stones.

But sometimes, you build something that truly matters—
something that changes how you see technology and its purpose.

For me, this is that project.

🔗 Project Repository:
https://github.com/MabelMoncy/TranscriberAppServerless
https://github.com/MabelMoncy/TranscriberAppWithServer

DEV Community: Mabel Anto Moncy