Turning a recording into useful text sounds simple until you try to make the experience feel fast, clear, and practical.
I recently built 123audio, a web app that converts audio and video files into editable transcripts with timestamps, speaker labels, summaries, translations, and subtitle exports.
Try it here: https://123audio.org
The goal was not to build another transcription dashboard full of settings. I wanted a lightweight workflow:
- Upload an audio or video file.
- Get a readable transcript.
- Review it alongside the audio.
- Turn it into notes, summaries, captions, or exportable documents.
This post covers the product decisions and technical approach behind it.
The problem
Audio is useful, but it is difficult to search, quote, reuse, and organize.
Meetings, customer calls, lectures, interviews, podcasts, and voice notes often contain valuable information that gets trapped inside a recording. Replaying an hour-long file to find one decision or quote is a pretty miserable workflow.
A transcript becomes much more useful when it includes:
- Punctuation and readable paragraphs
- Timestamps
- Speaker labels
- Searchable text
- AI summaries and action items
- Subtitle and document exports
That became the core product direction for 123audio.
The stack
The application is built with:
- Next.js 16 with the App Router
- React 19
- TypeScript
- Tailwind CSS
- Cloudflare for deployment and edge infrastructure
- Supabase for application data and authentication
- PayPal for billing
- A speech-to-text API for transcription
The frontend is intentionally browser-first. Users should be able to upload a file, receive a transcript, and start reviewing it without installing a desktop application.
Designing the transcription workflow
The first version had one important rule: the transcript should not feel like a dead text blob.
After upload, the interface keeps the original media and transcript close together. Users can play the audio, seek through it, and navigate the transcript through timestamps.
That matters because transcription is never perfect. Names, jargon, overlapping speakers, and poor audio quality still need human review. A useful transcription product needs to make correction and verification easy.
The UI is centered around:
- File upload
- Upload progress and processing state
- Audio playback controls
- Timestamped transcript blocks
- Speaker labels where available
- One-click AI actions
- Export options
Making transcripts useful after transcription
Raw text is only the first output.
Most people do not upload a recording because they want a long wall of text. They want an outcome: meeting notes, subtitles, a summary, a translated version, or a document they can share.
That is why 123audio includes actions to:
- Summarize a long transcript
- Extract decisions and action items
- Translate the transcript
- Rewrite spoken language into readable paragraphs
- Generate caption text
- Export as TXT, DOCX, PDF, SRT, or VTT
This changed how I thought about the product. The real job is not “audio to text.” It is helping someone get from spoken information to a usable artifact.
Handling privacy
Audio files can be sensitive. A meeting recording, interview, medical note, or customer call should not feel casually disposable.
The product communicates a simple privacy model:
- Uploads are encrypted
- Files are automatically deleted after 7 days
- User content is not used to train AI models
Privacy is not just a policy-page problem. It has to be visible in the product flow, especially at the moment someone decides whether to upload a file.
A product lesson: reduce the gap between upload and value
The key UX question was:
How quickly can a user go from “I have this recording” to “I can do something useful with it”?
Every extra step weakens that experience.
So the product avoids forcing users into a complicated setup before they can see value. A person can upload a file, get a transcript, review it, and decide later whether they need exports, saved history, or advanced features.
For a tool like this, the first successful transcript is the activation moment.
What I am still improving
There is plenty left to build:
- Better handling of noisy and multi-speaker audio
- More transcript editing tools
- Faster processing for large files
- Better organization for saved recordings
- More flexible export templates
- Clearer workflows for meetings, creators, students, and researchers
Transcription is a crowded category, but I think there is still room for products that feel focused, transparent, and genuinely useful after the transcript is generated.
You can try 123audio here:
I would love feedback from developers and builders, especially around transcription accuracy, privacy expectations, browser-based media workflows, and what would make a tool like this part of your regular workflow.
Top comments (0)