VICIdial audio store file formats

You upload a perfectly good MP3 and the audio store throws a red warning. Nothing is broken — VICIdial accepts exactly two formats, and an MP3 is not one of them. Get the format right and the warning never appears.

The two accepted formats are narrow on purpose. Asterisk plays prompts at telephone quality, and these specs match what the dial plan expects without any on-the-fly conversion that could glitch mid-call.

The only two that work

Here is the exact spec the audio store will accept:

• .wav — PCM mono, 16-bit, 8k. That is one channel, 8000 framerate, 16 bits, uncompressed PCM.
• .gsm — 8-bit, 8k.

Anything outside those two gets rejected. Stereo throws it off — telephony is mono, so a two-channel file is wrong even if it sounds fine on your laptop. A 44.1k or 48k sample rate is wrong; the line runs at 8000 Hz. A compressed Codec inside a .wav container is wrong; VICIdial wants raw PCM. After a .wav upload the page validates it, and the red warning is the store telling you one of those things is off. Re-format and upload again — a file with the same name simply replaces the bad version, so there is no cleanup to do.

The two extensions are not interchangeable in how they store the audio. A .wav holds the raw samples in a small header-plus-data wrapper. A .gsm holds audio already squeezed by a telephony compression scheme. Both arrive at the same 8 kHz playback, which is why VICIdial is happy with either — but only at the exact bit depth and rate above. There is no third option and no way to coax an MP3 or an AAC through; the store checks the format, not the file extension you typed.

Why 8k mono PCM

Phone calls do not carry hi-fi audio. The classic G.711 codec codec on a SIP (Session Initiation Protocol) trunk samples at 8000 Hz, one channel, so a prompt recorded to that exact spec plays back with no resampling and no quality loss in transit. Feed it a 48k stereo master and even if it converted, you would gain file size and lose nothing the caller can hear — the line cannot reproduce the extra detail anyway. Matching the format up front means the prompt sounds the same to the caller as it does in your audio chooser preview.

flowchart TD
  A["Source file"] --> B{"Container?"}
  B -->|".wav"| C{"PCM mono 16-bit 8k?"}
  B -->|".gsm"| D{"8-bit 8k?"}
  B -->|"mp3 / other"| R["Rejected: red warning"]
  C -->|"yes"| OK["Accepted"]
  C -->|"no"| R
  D -->|"yes"| OK
  D -->|"no"| R

Checking a file before you upload

You can verify a file before it ever touches the store. On Linux, soxi greeting.wav prints the channels, sample rate, and bit depth — you want 1 channel, 8000 Hz, 16-bit for a .wav. If your editor exported the wrong thing, convert it rather than fighting the validator. The difference between the two formats matters too: .gsm files are smaller, .wav files are higher quality, and which to choose depends on the prompt.

For the actual conversion commands, converting audio for VICIdial has the sox and ffmpeg recipes, and GSM vs WAV weighs size against quality. The wider audio prompts and voicemail guide puts both in context.

Format mistakes are the most common reason a prompt fails to load, and they are entirely avoidable once you know the spec. A managed VICIdial does not change the rules — they are baked into Asterisk — but it does hand you a clean, in-sync store so a correctly formatted file just works. If you would rather not run the plumbing, spin up a ready VICIdial server — see pricing. A box is live in under 40 seconds.