Online Tool Store
· 4 min read
How to Prepare On-Hold Audio for a Phone System
Manesh Jayawardhana
CIO & Co-founder
You record a clean, warm greeting for the phone system. It sounds excellent in the editor. You upload it, ring the number, and hear something thin, slightly too quiet, and — when the hold music loops — punctuated by an audible click every ninety seconds.
Telephone audio is its own medium with its own constraints, and almost none of the habits that make a podcast sound good apply to it.
What the phone network does to your audio
A traditional voice call carries roughly 300 Hz to 3,400 Hz. Everything below and above that is discarded before the caller hears anything. That’s the entire foundation gone from most music, and the air and sparkle gone from most voice recordings.
The sample rate follows from that band. Most PBX platforms — Asterisk and FreePBX among them — expect 8 kHz mono, often as 16-bit PCM WAV, sometimes as GSM or µ-law. Hand a system a 44.1 kHz file when it expects 8 kHz and, depending on how it handles the mismatch, playback can come out sped up and comical.
Then there’s looping. Hold music repeats, and any silence or DC offset at the file’s start and end becomes a click every time it wraps.
Why people get stuck here
- Sample rate mismatches. The single most common failure, and its symptom — chipmunk playback — looks like a broken file rather than a format issue.
- Mixing for headphones. A bass-heavy bed sounds full in the studio and disappears entirely through the voice band.
- Inconsistent levels between prompts. A greeting recorded loud followed by quiet hold music makes callers reach for the volume twice per call.
- Untrimmed loop points. A tenth of a second of silence at each end is inaudible once and irritating on the fiftieth loop.
What good telephony audio sounds like
Mixed for the midrange
Whatever survives 300–3,400 Hz is your whole mix. Voice sits comfortably there. Music needs to be chosen or arranged so its identity lives in that range rather than in bass or air.
Consistent loudness across every prompt
Pick a target and apply it to every file in the system — greeting, menu, hold, voicemail. Consistency matters more than absolute level, because the caller adjusts once and then stops noticing.
Loop-safe start and end
Trim leading and trailing silence to zero crossings and apply a short crossfade. A couple of hundred milliseconds is usually enough to make a loop seamless.
| Setting | Typical Target | Why |
|---|---|---|
| Sample rate | 8 kHz mono | What most PBX systems expect |
| Format | 16-bit PCM WAV, GSM or µ-law | Check your platform’s docs |
| Loudness | Consistent across all prompts | Callers adjust volume once |
| Loop | Trimmed, short crossfade | Removes the repeat click |
Common mistakes to avoid
- Uploading an MP3 to a system that expects WAV and assuming silence means the file was rejected — sometimes it plays as noise instead.
- Normalising to peak rather than to a loudness target, so a quiet-but-peaky recording ends up much softer than the music around it.
- Recording prompts in a room with audible air conditioning. The voice band strips warmth but keeps hum perfectly.
- Using licensed commercial music without checking the licence covers on-hold playback — it’s a public performance in many jurisdictions.
- Testing only through a softphone on a laptop, which often uses a wideband codec and hides every narrowband problem.
How to do it with IVR Audio Converter
The IVR Audio Converter sets up the conversion with telephony defaults, and the file stays on your device.
- Add the prompt or hold music you recorded.
- Choose the target format your platform expects — check its documentation rather than guessing.
- Set a loudness target and apply the same one to every prompt in the system.
- Enable loop trimming and crossfade for anything that repeats.
- Test by calling the number from an actual phone, not a desktop softphone.
Other browser-only audio utilities are in the tools directory.
Frequently asked questions
Why does hold music sound so thin on the phone?
Because the network discards everything outside roughly 300–3,400 Hz. The ITU-T G.711 standard that underpins traditional telephony encodes exactly that band. Choose music whose character survives in the midrange.
Which format does my system need?
Asterisk-based systems commonly want 8 kHz mono WAV or GSM. Hosted providers often accept MP3 and convert server-side. The provider’s documentation is the only reliable answer — formats vary more than you’d expect.
How loud should prompts be?
Aim for consistent rather than loud. Prompts noticeably louder than the hold music are more annoying than prompts that are slightly quiet, because the caller has to keep adjusting.
Final thought
Test on a real handset over a real call before you consider it done. Everything that goes wrong with telephony audio goes wrong specifically in the narrow band that a laptop preview doesn’t reproduce.