In development, not on Google Play yetTurns a room full of phones into one speaker: every phone plays the same song at the same moment, with no internet.
For the people using it
It started at a jam with friends in college. Nobody had a speaker, but everyone had a phone. Loudly turns those phones into one sound system: one person hosts, everyone else joins with a code, and every phone plays the same song at the same moment.
Why it's hard
Every phone has a different processor, scheduler and speaker delay, and ears hear 20 to 30 ms as an echo. Sonos, AirPlay and Chromecast solve this with their own hardware. Loudly does it in software, on ordinary Android phones, with no internet. I'm building it alone.
A gap of 20 to 30 ms between two phones already sounds like an echo.
Different processors, schedulers and speaker delays all pull the timing apart.
Discovery, timing and the music itself all have to travel phone to phone.
A connection can drop to Bluetooth or a guest can walk away, and the party has to carry on.
Steady on two phones on 8 October 2026, and in sync by ear. A microphone test is still to come.
Phones talk directly over Wi-Fi Direct or Bluetooth, picked automatically.
In one room, playing the same song.
One host and up to 14 guests. Scaling to it is the work now.
Host
Hosting is one tap; joining is a 6-character code.
One room, one speaker. The host creates a room; everyone else joins it. Rooms you've hosted are kept for next time.
Name it
A 6-character code is generated automatically.
Give the room a name friends will recognise. The code is made for you and the room is saved, so the next party starts faster.
Join
Wi-Fi Direct or Bluetooth, chosen automatically.
Guests type the code or pick the room from the nearby list. Phones connect to each other directly, with no internet needed.
Play
Live sync status for the room, and for every speaker.
The host picks a song from their library and controls play, pause, seek, volume and track for the whole room. A banner says whether everyone is in sync, and late joiners drop into the song mid-play.
Layer 1
Jetpack Compose and Material 3, with one ViewModel, state and intent per screen. Screens never touch audio or radios.
Layer 2
Pure Kotlin rules in the middle, and one sync engine that is the only place allowed to work out clock offsets.
Layer 3
Android's own decoders and low-level audio output, with Opus over a peer-to-peer link.
Everything happens phone to phone. There is no server anywhere in the loop.
Only the sync engine works out clock offsets and drift. Screens never touch audio, codecs or radios, so timing bugs have one place to live.
When a guest drops to a slow Bluetooth link, the room says so, the sync targets relax and the stream bitrate falls from 128 to 48 kbps.
128 → 48 kbps on slow linksA guest that loses the host tries three times, four seconds apart. The host keeps a dropped guest's place for 15 seconds, and a returning guest rejoins mid-song.
Changes are checked on real phones with the app's own timing readings, and by ear. A microphone recording of the phones playing together is the next check.
Loudly isn't on Google Play yet. On 8 October 2026, on two phones, the app's own timing reading settled at about 11 ms and the music was in sync by ear. A microphone test on two and three phones is next. The screens here are from a test build on an emulator.
Audio, timing and live updates where milliseconds matter, measured on real devices.
Peer-to-peer features that keep going with no internet and no server.
Kotlin from Compose screens down to MediaCodec and AudioTrack, with coroutines and Flow throughout.
Someone who checks on real devices before claiming a result.