Can I code this? A Voice Memo App With Transcription
Recording audio and sending it to an existing speech-to-text API gets you a real transcript quickly — that part is thoroughly solved. It's the same well-bounded core as Otter.ai, just aimed at quick personal notes instead of meetings.
Borderline
Effort: a weekendRunning cost: $0/mo to start
Why CanICodeThis says Borderline
Recording audio from the browser's or OS's native microphone API is well-documented, mature work
Sending the recording to an existing speech-to-text API and displaying the transcript is a bounded integration
Organizing memos into a searchable list with tags is ordinary CRUD
The parts that'll cause problems
Real-time transcription while still recording (not after-the-fact) needs streaming audio infrastructure, a genuinely bigger step
Speaker diarization if you record conversations, not just yourself, adds real complexity most transcription APIs handle with varying accuracy
Offline transcription (no internet needed) would mean running a model locally, a different and heavier undertaking
How I'd build this
frontend
Next.js
database
Supabase
hosting
Vercel
MVP scope
Record audio from the microphone
Send to an existing speech-to-text API for transcription
Store memos with searchable transcripts and tags
Postpone to v2
Real-time streaming transcription while recording
Speaker diarization for multi-person recordings
Offline/local transcription
What it actually costs to build
AI agent — with a subscription (Claude Pro/Max, Cursor, etc.)
$0 marginal
AI agent — pay-per-use API, no subscription
$43–$86 one-time
Hosting, once it's running
$0/mo (free tier)
Domain name, if you want your own
~$12/yr
Existing tools solving a similar problem typically run $16.99–$16.99/mo — see how they compare below.
The build prompt
Genuinely one of the simpler builds here:
1. Record audio from the microphone using the browser's native media APIs.
2. Upload it and send it to an existing speech-to-text API for transcription.
3. Store memos in a searchable list with tags.
After-the-fact transcription (not live) is a reasonable, achievable starting point. Out of scope: real-time streaming transcription while recording, which needs streaming audio infrastructure, a bigger step than after-the-fact processing, and speaker diarization for multi-person recordings, a genuinely hard audio-ML problem.