The title never changes
The station isn’t sending updates, or someone typed a fixed slogan into the title field and left it there.
Metadata is the small amount of text that travels with a radio stream: the station’s name, the song that’s on, sometimes the program and a picture. It’s the reason one station fills your screen with artwork and the next one shows a name and nothing else. Here’s where it comes from and what a player does with it.
The audio is the sound. Metadata is everything written about it: the station name, the current song, the program, the codec and bitrate, the genre, the language, and any artwork. A player can’t hear a song title. It has to be told.
There are two kinds. The first describes the station and rarely changes: name, country, language, website, tags, logo. The second changes while you listen: the now playing line, the program name, sometimes a flag that says an ad break is on.
It starts inside the station. Somewhere between the studio and the stream there’s playout software that knows which song is on, an encoder that packs the audio, a server that sends it out, and occasionally a separate web service with program details. Each of those can add or lose a piece of the story.
The encoder takes the live audio and pushes it to the server. On the way it stamps the stream with the basics: station name, genre, a short description, the website, the bitrate, and the codec.
Icecast, SHOUTcast, and the commercial platforms built on the same ideas receive that stream and hand it to listeners. They’re also where the metadata becomes readable from the outside, so this is the layer a player actually talks to.
The song title usually comes from the playout system, the software that runs the schedule: RadioDJ, mAirList, AzuraCast, Rivendell, Zetta, and plenty of others. When a track starts, it tells the server. If that link isn’t set up, the server never learns what’s playing, and neither do you.
The station name is who’s broadcasting. The stream title is the line everyone looks at, because that’s where the current song or program appears. Genre and tags are the loose labels that say “jazz” or “talk” or “news”.
Bitrate, in kbps, is a rough guide to sound quality and to how much data you’ll use. The codec is how the audio is compressed: MP3, AAC, AAC+, Opus, Ogg Vorbis, occasionally FLAC. Sample rate, usually 44.1 or 48 kHz, is how finely the sound was measured before it was encoded. You can safely ignore that last one.
Artwork might be the station logo, the album cover for the current song, or a picture for the program. The description, website, language, and country fields are what let a directory sort thousands of stations into something you can browse.
Most online stations send their now playing text the same way, using a trick that SHOUTcast invented in the late 1990s and everyone else copied. It’s called ICY, and nearly every radio app, desktop player, and server still speaks it.
The idea is simple. The player asks for metadata by sending an Icy-MetaData: 1 header. The
server replies with the audio and, every fixed number of bytes (announced as icy-metaint),
squeezes in a short block of text. The player pulls the text out, plays the audio, and carries on.
At the start of the stream you get the station-level headers: icy-name, icy-genre,
icy-br, icy-url, icy-description, and icy-pub. In the
blocks that follow, StreamTitle is the one that matters. That’s the now playing line.
It’s simple, it costs nothing, and it works with the servers everyone already runs. One wrinkle: a web page can’t always read those headers directly because of browser security rules, which is why the WorldTune web player fetches them through its own backend rather than straight from the station.
ICY is still the default, but bigger broadcasters, especially ones with video infrastructure or their own apps, have other options. These tend to line up better with the audio when the stream is delivered in segments, and they can carry more than a single line of text.
When audio is cut into HLS segments, the metadata can ride inside them as ID3 tags, or sit in the playlist as
EXT-X-DATERANGE entries. Either way it arrives with a timestamp, so the title changes exactly when
the song does.
DASH does something similar with events listed in its manifest. You’ll rarely meet it on a pure radio stream, but broadcasters who already use it for video sometimes use it for audio too.
Some stations run a small web service that answers “what’s playing?” with the current track, the last few songs, the program name, and artwork. A player can read that alongside the stream and show far more than a single ICY line could.
Stations that also broadcast on FM or DAB+ already send song text to car radios through RDS or DLS. Some forward the same feed to their online stream, so every platform shows the same thing.
The codec isn’t metadata, but it hints at how the metadata will arrive. An MP3 stream almost always uses ICY. AAC and AAC+ streams use ICY too, unless they’ve been wrapped in HLS, in which case the text usually comes as ID3 tags inside the segments.
Ogg Vorbis and Opus can carry their own tags inside the container, and Icecast knows how to update them, though not every player reads them. FLAC is rare on live radio because of the bandwidth, but when it does appear it can carry the richest tags of all.
Open a few stations in the web player and watch the now playing line. Some will change with every song and pull in artwork. Some will show the program name. Some will show the station name and nothing else, and now you know why.
It comes down to how the station is wired up. Great audio and no metadata is common, and it usually means one link in the chain, often between the playout software and the encoder, was never connected.
If the software that plays the songs never tells the encoder what’s on, the stream only ever carries the station name. Small, volunteer-run, and live-heavy stations are the usual suspects, and often nobody has noticed.
During news, sport, or a live presenter there’s no song to report. Some stations send the program name, some send nothing, and some leave the last song’s title sitting there for an hour.
A station that rebroadcasts another network often passes the audio through but not the text. You end up seeing the relay’s name instead of what’s actually on.
Ads and jingles usually aren’t sent as tracks. Some systems flag them as advertising. Others just leave the previous title up until the music comes back.
Three places. The station logo comes from the broadcaster or the directory and never changes. Album art is usually looked up by the player, using the artist and title from the stream to search a music database. Program art, when it exists, comes from the station’s own web service.
When the lookup fails, because the title was misspelled or the song is too obscure, a good player falls back to the program art or the logo. Seeing a logo instead of a cover isn’t a bug. It’s the player being honest about what it could find.
Reading the text is the easy part. The hard part is being graceful when the text is missing, late, garbled, or wrong, which happens more often than you’d think.
The player opens the stream, asks for ICY metadata, and starts receiving audio with text blocks mixed in. For HLS or DASH it reads the tags from the segments or the playlist instead, and sometimes asks a separate now playing service as well.
It pulls out StreamTitle, tidies it up, and tries to split it into artist and title when the
format is obviously Artist - Title. It also has to accept that the “title” might be a show
name, a slogan, or the station’s phone number, and not pretend those are songs.
With an artist and a title in hand, the player can search outside services for a cover, a short bio, or the lyrics. That part is optional, it fails on ambiguous names, and every app does it a little differently.
A good player falls back to the station name, doesn’t flicker when a field updates twice in a second, and quietly fixes accented letters that arrive in the wrong text encoding.
Almost all of these start at the station. A player can tidy up what it receives, but it can’t invent a song title the broadcaster never sent.
The station isn’t sending updates, or someone typed a fixed slogan into the title field and left it there.
The playout system is running ahead or behind the audio, or a relay is passing on stale text.
The station sent the text in one encoding and declared another. Classic Latin-1 versus UTF-8.
The lookup found nothing, the title was too vague, or what’s playing is a talk show with no album to show.
ICY updates arrive every so many bytes, not every so many seconds, so on a low-bitrate stream they come less often.
Talk, news, and sports stations tend to send the program name rather than track-by-track details, because there aren’t any.
The text in a stream belongs to the broadcaster, and WorldTune shows it as sent. We don’t guess at song titles, invent program names, or add rights information the station didn’t provide.
Cover art and artist details come from public music databases, matched on the artist and title the station gave us. Whether a stream is online, where it can be heard, and how accurate its text is remains the broadcaster’s call.
A radio stream is two things travelling together: the sound, and a note about the sound. The station decides how good each one is, separately. Once you see it that way, most of what a player shows (or doesn’t) makes sense: the title that updates, the cover that appears, and the station that only ever shows its name.
For the practical side of finding and saving stations, there’s the listening guide. For how we got here in the first place, there’s the history of online radio.