How to convert audio for VICIdial

You recorded a greeting in a normal audio editor, exported it, and the VICIdial audio store rejected it. The source is almost always stereo, the wrong sample rate, or a compressed codec. One command fixes all three and hands you a file the store accepts on the first try.

The target is fixed: PCM mono, 16-bit, 8000 Hz for a .wav, or 8-bit 8k for a .gsm. Every flag below exists to hit that target exactly.

Convert with sox

sox is the cleanest tool for this. To turn any input into a VICIdial-ready WAV:

sox input.mp3 -r 8000 -c 1 -b 16 -e signed-integer output.wav

Reading the flags: -r 8000 sets the sample rate to 8k to match the phone line, -c 1 forces mono (one channel), -b 16 sets 16-bit depth, and -e signed-integer makes it PCM rather than a compressed Codec. That is the full PCM mono 16-bit 8k spec. For a .gsm instead, sox input.wav -r 8000 -c 1 output.gsm produces the 8k GSM file.

Convert with ffmpeg

If ffmpeg is what you have, the equivalent is:

ffmpeg -i input.mp3 -ar 8000 -ac 1 -acodec pcm_s16le output.wav

Here -ar 8000 is the sample rate, -ac 1 is mono, and -acodec pcm_s16le is signed 16-bit little-endian PCM — the same uncompressed format sox produces. Both tools land in the same place; pick whichever is already installed. ffmpeg is the safer bet for odd input containers because it reads almost anything, while sox gives you cleaner resampling control if you are fussy about quality. For a single greeting either is fine.

If the source is loud, distorted, or recorded too quiet, fix the level before you downsample, not after. sox can normalize in the same pass — adding gain -n to the sox command levels the volume so the prompt is not jarring against your hold music or an IVR (interactive voice response) menu. Trim dead air off the front and back too; a half-second of silence before a greeting feels like a dropped call to the person on the line.

flowchart LR
  A["Source: mp3 / 48k stereo"] --> B["Resample to 8000 Hz"]
  B --> C["Downmix to mono"]
  C --> D["Encode 16-bit PCM"]
  D --> E["Verify with soxi"]
  E --> F["Upload to audio store"]

Verify, then upload

Do not trust the conversion blind — check it. soxi output.wav prints the channels, sample rate, and bit depth, and you want to see exactly 1 channel, 8000 Hz, 16-bit. A quick checklist before upload:

• Channels: 1 (mono). Stereo is rejected.
• Sample rate: 8000 Hz. Not 44100 or 48000.
• Encoding: signed 16-bit PCM for .wav, or 8-bit for .gsm.

With that confirmed, the upload validates without a red warning and the file distributes to every server. The same recipe works whether the prompt is a Welcome message, hold Music on hold, or a clip for a Soundboard — they all need the same format. If you batch-convert a folder of clips, run the command in a loop and verify each one before pushing them in.

Once a file is converted and verified, uploading the audio prompt is the next step, and the broader audio prompts and voicemail guide shows where the finished prompt gets used.

Conversion is a one-liner, but it is one more thing to remember when you are setting up a dialer from scratch. A managed VICIdial gives you a clean, in-sync audio store so the only step left is the convert-and-upload above. If skipping the rest of the build sounds good, spin up a ready VICIdial server — see pricing. It is provisioned in under 40 seconds.