Plain explanations of the concepts behind an automated radio station: clock templates, cue points, the top of the hour, AI presentation, stream quality and what it all costs. Written from how NexRadio does it, useful for any system.
A clock template is the blueprint of one hour of radio: which kind of element goes where. The automation fills the clock with real tracks every hour. Understand the clock and you understand the station.
At HH:00:00 the new hour must start, but a song is never exactly finished. Two schools: the clock is sacred, or the song is sacred. Both are right, but you have to choose, and know what you chose.
Tight transitions are not a matter of taste but of measurement. Every track carries a handful of timestamps, and the automation calculates with them. Understand them and you hear why one station flows and another stumbles.
An AI presenter is not a voice reading a text. It is a chain: a template in the clock, a script written at the last moment with what is really playing, a render with the right voice, and placement to the millisecond.
The fear when switching is always the same: losing years of cue points and categories. It does not have to be. What can and cannot come along, and in which order to do it.
A good stream starts at the source file and ends in the listener's phone. Every step in between can cost quality. Where we touch nothing, where we process once, and why.
The licence is the visible part. The server, the backup, the stream, the voices, the website, the app and the hours someone spends on it are the rest. How to make a fair comparison.
The stream is the product; the app and the website are the shop. What belongs in a good listener app, why the playlist should be public, and how a request travels from the phone into the clock.
Starting a radio station in 2026 is no longer about equipment but about choices: music, format, automation, stream, rights and audience. The order in which you make them decides whether you are still on air after three months.
An AI presenter costs no salary but does cost spoken minutes. How much depends on how much you talk, which voice you choose and how well the automation keeps usage in check. The sum, with real numbers.
For years the automation sat on a PC in the studio. Now it can run on a server you never see. What you gain, what you give up and which questions to ask before you switch.
The things that make a station feel alive, news on the hour, the weather right now, a listener's request, are exactly the things that are hardest to automate. This is how we do it, and where it goes wrong.