IVR Audio File Formats Explained: How to Convert Any Recording for Your Phone System

It is one of the most common small frustrations in setting up a business phone system. Someone records a professional greeting — on a phone, in a studio, or with a text-to-speech tool — uploads it, and the system either refuses the file or plays it back sounding like it was recorded underwater.

The file is not broken. It is simply the wrong shape for a telephone network, which has requirements that nothing else in modern audio still has.

Why Phone Audio Is Different

The telephone network was standardised around the human voice, and only the human voice. The decision made decades ago — and still in force today — was to carry audio sampled 8,000 times per second, in a single channel, using a specific compression scheme. That is enough to make speech clearly intelligible and nothing more.

Music, by contrast, is normally 44,100 samples per second in stereo. When you hand a phone system a music-shaped file, one of two things happens: it rejects it, or it converts it badly on the fly. The second outcome is worse, because it sounds wrong and nobody can tell why.

So the target format for almost every IVR prompt, greeting, voicemail message and hold announcement is:

• WAV container — uncompressed, universally understood by phone systems
• 8,000 Hz sample rate — the telephone standard
• Mono — one channel; a phone call has no stereo
• 16-bit PCM, or in some systems 8-bit µ-law

A file matching that description will load into essentially any PBX, including WOCOM Cloud PBX, Asterisk, FreePBX, 3CX and most legacy systems.

What Each Setting Actually Does

Sample rate is how many times per second the audio was measured. Higher captures more detail, particularly high frequencies. A phone line discards everything above roughly 3.4kHz regardless, so a 44.1kHz file carries a great deal of information that will be thrown away — and if the discarding is done badly, it produces the metallic artefacts people describe as “robotic”.

Channels matter for a different reason. If your recording is stereo and the system takes only the left channel, and the voice happened to be recorded predominantly on the right, the result is a very quiet or empty prompt. This is a surprisingly common cause of “the greeting uploaded but plays silence”.

Bit depth is the resolution of each measurement. 16-bit is standard and safe. Some older systems want 8-bit µ-law, which is what the telephone network itself uses.

Level is not a format setting but causes as many problems as one. A greeting recorded too quietly is unintelligible on a mobile in a noisy environment; one recorded too loudly clips and distorts. Aim for a strong, even level without peaks hitting the ceiling.

Converting a File Without Installing Anything

WOCOM publishes a free audio converter for exactly this job. It runs entirely in your browser — the file is never uploaded to a server, which matters if the recording contains anything you would rather not send across the internet. Drop in an MP3, M4A, WAV or most other common formats and it produces a telephone-ready 8kHz mono WAV.

It is free to use whether or not you are a WOCOM customer, and it works for any phone system, not just ours.

Practical Advice on Recording the Prompt Itself

Format is the easy half. The recording matters more.

• Record in the quietest room you have, away from air conditioning, traffic and fans. Background noise that is barely noticeable in the room becomes very obvious over a phone line.
• Use a real microphone or a decent headset rather than speakerphone. A modern mobile held close to the mouth is better than a laptop microphone across a desk.
• Leave a short silence at the start and end. A prompt that begins the instant the file starts sounds clipped, because the first syllable is often cut by the system.
• Speak slightly slower than feels natural. Phone audio is compressed and callers are often in noisy places.
• Keep menu prompts short. Lead with the number: “For sales, press 1”. The reasons are covered in why customers hang up on phone menus.

Hold Music Has Its Own Rules

Hold music runs through the same 8kHz mono pipe, which is why so much of it sounds terrible. Two consequences follow. First, choose material that survives the treatment — sparse, mid-range, without heavy bass or cymbals. Second, and more importantly, check that you have the right to play it. Commercial music generally requires a licence for use on a business phone system, and a great many Jamaican businesses are playing radio recordings they have no licence for. Purpose-licensed hold music, or a short recorded message loop, avoids the question entirely. On-hold messages goes into what to say.

If You Would Rather Not Record It Yourself

Two alternatives are worth knowing about. Modern text-to-speech is genuinely good now and produces clean, consistent prompts you can regenerate whenever your hours or departments change — no re-recording session required. And if you run the WOCOM AI receptionist, the question of prompt recording largely disappears: Alex speaks naturally, adapts to what the caller asks, and does not need a new WAV file every time something changes.

For help getting prompts loaded onto your system, call 876-326-1212 or message us on WhatsApp. The converter is free and needs no account.