The encoder says it is connected. The Icecast status page lists your mount, the player’s timer keeps counting, and there is nothing to hear. Every part of the system that reports a status is reporting success, which makes this failure confusing the first time you run into it.
It also narrows the search. If a player connects and keeps receiving data, Icecast and the network between it and the listener are doing their job. The audio is either already missing when the encoder picks it up, or it is in the stream and something on the listening side drops it. This post works through both halves in the order that usually finds the cause fastest. The Icecast and IceS details come from the official Icecast documentation, the Icecast FAQ and the IceS FAQ. The measurement commands use filters from the FFmpeg filter documentation.
Why Icecast doesn’t notice
Icecast passes on whatever the source client sends. It doesn’t decode the audio or measure its level. The closest thing it has to a health check is <source-timeout>, which removes a source that sends no data for that many seconds (the sample config uses 10). An encoder with a silent input keeps encoding that silence and sending it, so it never hits the timeout and the mount stays up.
We went through the stats fields one by one in Icecast status-json.xsl can look healthy while listeners hear silence. For troubleshooting, it’s enough to know that a green status page rules out nothing that happens before the encoder.
Step 1: Is the silence in the stream or in your player?
Start here, because it splits the problem in half in a couple of minutes.
Open the public listener URL, the one your listeners actually use, in a second player on a different device. VLC or a browser tab is fine. The Icecast setup guide gives the format as http://host:port/mountpoint, with the example http://192.0.2.23:8000/mystream.ogg. Adding .m3u to that URL returns a small playlist file that points at the stream. For this test, open the stream URL itself so you know exactly which mount you are hearing.
Then measure the level instead of trusting your ears and whatever speakers are nearby. FFmpeg’s volumedetect filter prints the mean and maximum volume of the audio it decoded:
ffmpeg -hide_banner -nostats -t 30 \
-i https://stream.example.com/live \
-af volumedetect -f null -
volumedetect only prints its numbers when the input ends, so a live stream needs the -t 30 limit. Look for mean_volume and max_volume in the output. Both are in dB relative to the maximum sample value. There is no single number that means “silent” for every station, so run the same command once on a normal day and save the result. If max_volume during the outage is far below that reference, the stream really carries no programme audio.
If the second player has sound and the numbers look like your normal day, the stream is fine and the first player is the problem. The IceS FAQ says that if the stream gets to the player, the problem is in how the player handles it. Things to check:
- The player, the browser tab or the operating system is muted, or the audio is going to a different output device.
- The player can’t decode the format. The Icecast FAQ lists Ogg (Vorbis, Theora), Opus, FLAC and WebM as supported and says MP3, AAC and other non-free formats “might work” without official support.
- The player guesses the format from the file extension. Both the Icecast setup guide and the IceS FAQ mention players that need
.oggat the end of an Ogg Vorbis mountpoint.
If both players are silent and volumedetect agrees, the silence is in the stream. Move upstream.
Step 2: Is the right source on the mount?
Before you start pulling cables in the studio, confirm that the encoder you are looking at is the one your listeners hear. Icecast creates a mountpoint automatically when a source connects to it, so a typo in the encoder’s mount setting quietly creates a second mount. Your listeners stay on the first one, which might be held by another source or by nothing at all.
Open the stats (/status-json.xsl, or /admin/stats.xsl with the admin login) and check three things:
- Is there a second mount with almost the same name as your listener mount?
- Do
source_ipanduser_agenton the listener mount match the machine and encoder you expect? An old automation PC or a backup encoder that connected first will hold the mount. - Are your listeners on a fallback mount?
<fallback-mount>moves clients when the source shuts down, and<fallback-override>is the setting that lets a reconnecting source take them back. If the fallback is a silent file or a dead relay, listeners hear nothing while the main mount looks healthy.
Step 3: Check what the encoder is reading
If the right encoder holds the right mount and the stream is silent, the encoder is receiving silence. Most encoders have an input meter. A flat meter means the problem sits before the encoder.
The IceS FAQ has an entry that matches this symptom almost word for word: “My player connects but I don’t hear anything.” Its answer is that if the player gets data and shows the sample rate and channels, the mixer settings are probably wrong and IceS can only read silence. It mentions that ALSA may have many mixer levels, all muted by default. Another entry, on encoding from line-in, explains that where the sound card takes its audio from depends on the mixer settings, and that the default is usually the microphone. The FAQ suggests a tool such as alsamixer to see and change the capture device.
The same check applies to other encoders and operating systems. Look at which input device the encoder is set to and whether that device is still the one you think it is. A USB interface that was unplugged, a driver update or a reboot are good reasons to look twice.
Step 4: Follow the signal back through the mixer and playout
With a flat encoder meter, work backwards from the encoder toward the studio:
- the cable or virtual audio device between the mixer output and the encoder input
- the output or bus that feeds the stream, which may be an aux or monitor send rather than the main programme bus, and someone may have turned it down
- the faders and mute buttons on the channels routed to that bus
- the playout or automation software, which may have stopped at the end of a playlist
If Liquidsoap sits between playout and Icecast, keep mksafe in mind. It fills gaps with blank audio so the output never fails, which means a dropped input turns into silence on a connected mount. We covered that setup in AzuraCast and Liquidsoap fallbacks vs external silence monitoring.
Quick reference
| What you see | Where to look first |
|---|---|
| A second player has sound, the first doesn’t | The first player: mute, output device, format support, mount extension |
| Both players silent, encoder input meter flat | Encoder input device, capture mixer settings, cable, mixer bus, playout |
| Both players silent, encoder input meter moving | Mount name in the encoder, which source holds the listener mount, fallback mount |
| Main mount shows a healthy source but almost no listeners | Listeners stuck on the fallback mount; check fallback-override |
Hearing about it before a listener does
Once the audio is back, set up a check that listens to the stream, because the status page won’t warn you next time either. FFmpeg’s silencedetect filter logs when the level stays at or below a noise tolerance for at least a minimum duration. Its defaults are -60 dB and 2 seconds. A one-minute check on the public URL looks like this:
ffmpeg -hide_banner -nostats -t 60 \
-i https://stream.example.com/live \
-af silencedetect=noise=-50dB:d=10 -f null -
It prints silence_start and silence_end lines for each silent stretch. Treat -50 dB and 10 seconds as a starting point. Talk radio with long pauses needs different values than a music station, and our post on silence detection thresholds without false alarms goes through how to choose them. Run the check from a machine outside the studio and the streaming server, so it still works when either of them is the thing that broke.
SilenceAlarm does this as a hosted service. You add the stream URL your listeners use (Icecast, Shoutcast, HLS or another HTTP webstream), and it alerts you by email, mobile push, SMS or webhook, depending on the plan, when the stream goes silent or offline. The free plan covers one monitor with 15-minute check intervals. Pro lets you pick intervals from 1 to 60 minutes and set the level threshold and the length of each audio sample. If you just want to know whether a stream URL carries audio right now, the SilenceAlarm homepage has a free 20-second test that needs no signup.
FAQ
Why does Icecast show my stream as online when there’s no sound?
Icecast only checks whether the source keeps sending data. Encoded silence is still data, so the source stays connected and the mount looks normal.
How do I tell whether the problem is my player or the stream?
Play the public stream URL on a second device and measure it with FFmpeg’s volumedetect. If the second device has sound and the levels match a normal day, the problem is the first player. If both are silent, the silence is in the stream.
My encoder’s meter moves, but listeners hear nothing. What should I check?
Check the mount name in the encoder against the URL listeners use, look at source_ip and user_agent on that mount in the stats, and see whether listeners are sitting on a fallback mount.
Can Icecast alert me when the stream goes silent?
The Icecast configuration documentation has no setting that measures audio level. <source-timeout> only reacts when a source stops sending data. For silence you need a separate check that decodes the audio, such as an FFmpeg script on another machine or an external monitoring service.
1 Comment