Rule one: do not touch the source file
Music comes in as FLAC or WAV. We convert FLAC to WAV 24-bit purely as a format change, without normalisation, compression or limiting. Never from 24 to 16 bit. Those treatments belong in one place, in the transmitter, not at import, because everything you do at import is irreversible and stacks on top of what the transmitter does later.
One processing point
After the mixer comes one chain of processing: automatic gain control, parallel compression, soft clip, limiter, a five-band master bus, a true-peak limiter and a stereo widener. The chain has presets, from neutral to hit radio, that you can switch on air without a restart. If you are used to a separate processor such as StereoTool: it can stay on the stream if you want, but the mixer itself does no second treatment.
Why voice does get a separate step
A rendered AI voice arrives raw, quiet and uneven. The automatic gain control in the transmitter is deliberately slow, tens of seconds, and too slow for an eight-second link. So voice is brought to loudness before the mixer with high-pass, normalisation, compression and limiter. That is a creative render step, not a volume correction during the broadcast. Music does not get that step.
Output: MP3 320 and an HLS ladder
The main stream is MP3 at 320 kbps, the highest in common use. Alongside it a station delivers HLS in a ladder from 96 to 320 kbps, so a phone on a shaky network does not stall but steps down a rung. The player on the website and the app choose for themselves. Mind the details: the first seconds of a stream (the burst) decide whether a player starts quickly, and not every browser plays HLS equally well; we measure that per browser instead of assuming it.
What you hear
No music sagging under a voice, no pumping compressor, no jingle being faded. A song that runs at full strength to its own fade-out, a voice that arrives at the right loudness, and a stream that sounds like the source file with a consistent station colour over it.
Frequently asked questions
Isn't 320 kbps MP3 old-fashioned? Why not AAC?
MP3 320 plays everywhere, from old car radios to every browser, and is transparent at that bitrate. HLS provides the mobile variants. AAC variants can be added on request.
Can I use my own StereoTool setting?
Yes, on the stream. The mixer then does no processing of its own on top; the choice is per station.
What is an HLS ladder?
Several qualities of the same stream side by side, from 96 to 320 kbps, in short segments. The player switches automatically to what the network can handle at that moment.