Why meditation is about audio plumbing, not AI

A meditation app is 80% audio plumbing and 20% UI. If the audio does not keep playing when the phone locks, your app is broken regardless of how good the content is. AI is the easy part here — generating session scripts, translating them, picking the next session for the user. The hard part is background playback on iOS and Android.

I have not shipped a meditation app myself, but I have shipped audio playback inside one of the four apps I built, and the lessons carry over cleanly.

What you actually need to build

• Session library: List of guided meditations with length, category, and narrator. A single sessions table with a storage URL works for v1.
• Audio player: Play, pause, seek, background playback, lock-screen controls. This is the part where AI code generation usually fails silently.
• Streaks: Days meditated in a row. One session of any length counts. Keep the rule simple.
• Daily bell: A scheduled local notification at a user-chosen time. "Two minutes is enough" beats "perfect session at 6 a.m."
• AI session picker: Given how the user has been using the app, suggest the next session. One call per session start is fine. Do not call the LLM during playback.

Skip in v1: journaling, sleep stories, breathing exercises, community feed. Add those once the audio layer is stable.

The stack I use

• React Native + Expo.
• expo-av for audio playback plus expo-notifications for the daily bell.
• For background playback: enable the iOS background audio mode in app.json and wire react-native-track-player if you need richer lock-screen controls.
• Supabase — auth, session metadata, streak table. Supabase Storage for the audio files themselves (cheaper than a CDN until you are big).
• NestJS — the AI session picker endpoint.

Real build time

With the boilerplate and a small session library, meditation apps land around 3 weekends.

• Auth + session list + detail: ~6 hours.
• Audio player + background mode + lock screen: ~10 hours. Budget more if you have never done this.
• Streaks + daily bell: ~5 hours.
• AI session picker: ~4 hours.
• Store submission: ~5 hours.

About 30 hours. Audio is always the biggest block.

Where people get stuck

• Audio stops when the screen locks: You forgot to set staysActiveInBackground: true on the Audio mode, and you did not enable the audio background mode in app.json. Both are required on iOS.
• Android audio focus: On Android you need to request audio focus and pause when the user takes a call. expo-av handles most of this but you still need to test it on a real device, not a simulator.
• Audio files too big: 64 kbps mono MP3 is fine for voice. Do not ship 320 kbps stereo. File size directly hits your Supabase Storage bill and the user download time.

Skip the setup

Audio player scaffold, background mode config, Supabase schema, rate-limited AI session picker — pre-wired. The 11 AI agents scaffold the session list and player screens.

See pricing