The sound is the DJ's own mix

SONORA takes its master output at one point: after the channels, the crossfader, the master effects and the limiter, before the sound card. Headphone cueing is a separate path and is never sent. That signal is encoded as stereo Opus (256 kb/s unless the DJ picks otherwise) in a background worker, so the DJ's own playback never waits for the network, and sent to this server, which passes the same packets to every listener.

Opus is a high-quality lossy codec, not lossless, and listeners don't receive SONORA's internal 32-bit float mix: they receive a very good 48 kHz stereo copy of it.

The booth on the page is drawn, not filmed

There is no screen capture. SONORA sends the state its engine has applied (each knob, fader, pad, loop and effect, whether moved with the mouse, a controller or automation) and the decks' positions many times a second. The page draws the decks, waveforms and mixer from that. Controls on the page show values; listeners can't change the mix.

Why what you see matches what you hear

State reaches your browser before the audio it describes: the audio sits in a short buffer that smooths out the network. Every audio packet carries the exact moment of SONORA's audio clock it came from, and your browser schedules it itself, so the page knows which moment of the DJ's mix is leaving your speakers right now. The booth is drawn at that moment from a short history of the desk, not at the newest state. A filter sweep on screen moves with the sweep you hear.

The status line in a room says how sure it is: Locked to audio means the page is following the audio clock; Estimated means it is using a guess (older browsers that need the backup MP3 stream); Preview means you haven't pressed Listen yet, so the booth runs at an assumed delay.

Headphones or speakers over Bluetooth add their own delay, which browsers can't always measure.

Delay

Listeners usually hear a set between a third of a second and a second after the DJ, depending on their connection. If the connection hiccups, the buffer grows a little so the sound doesn't break up.

Privacy

DJs choose whether tracks waiting on a deck are shown before they are heard, and whether hot cues and waveforms are shared. Hidden details aren't sent to listeners at all, rather than being hidden on the page. Played-track lists only count a track once it has been audible for half a minute.

Back to back

Two SONORA DJs can play one desk from two places (two decks, one each, or four, two each). Each machine plays the whole desk from the same files; only moves travel between them, stamped with a shared clock. One of the two broadcasts the result here.