No audio on VICIdial calls: first things to check

The call connects, the agent and the lead are both on the line, and there is dead silence in both directions. The signalling worked, so the problem is almost never SIP. It is the media. Voice rides on RTP, the Real-time Transport Protocol that carries the actual audio packets separately from the call setup. When you hear nothing both ways, RTP packets are not reaching the box, leaving it, or both.

In VICIdial, the SIP (Session Initiation Protocol) handshake and the RTP stream are two different conversations on different ports. A call can ring, answer, and time perfectly while no audio ever moves. That split is the whole reason silent calls are confusing.

What's actually happening

Signalling negotiates the call on one port. Each side then advertises, in the call setup, an IP and port where it expects audio. If the box advertises a private IP, or a firewall drops the RTP range, the packets go nowhere and you get silence. Codec mismatch causes the same symptom: both ends agree to talk but cannot decode what arrives. Because the agent leg and the carrier leg are negotiated independently, you can have audio on one and silence on the other, which is the difference between this symptom and one-way audio.

The fastest confirmation is to watch the RTP counters during a live silent call. If the box reports zero received packets, audio is not arriving and the cause is inbound: a closed port, a private advertised IP, or a carrier holding media. If it reports received packets but sends none, the problem is on the agent side. That single observation cuts the search in half before you change anything.

flowchart TD
  A[Call answers] --> B{RTP ports open}
  B -- No --> S[Silence both ways]
  B -- Yes --> C{External IP correct}
  C -- No --> S
  C -- Yes --> D{Codec agreed}
  D -- No --> S
  D -- Yes --> E{Carrier sends media}
  E -- No --> S
  E -- Yes --> F[Audio flows]

Check these in order

1. RTP port range open. Asterisk uses a UDP range (commonly 10000-20000). Confirm those ports are open inbound and outbound on the firewall and any upstream network device. A single direction open gives one-way audio, not total silence, but a fully closed range gives silence both ways.
2. External IP correct. On a public box behind any address translation, the external IP and local network settings in the SIP config must match the box's real public address. If it advertises a private IP, the far end sends audio into a void. NAT handling, the NAT traversal settings, is the most common culprit on a misconfigured install.
3. Codec match. The agent leg and the carrier leg must share a codec. If the box offers only one Codec and the carrier offers another, the call connects but neither side can render the other's packets. Allow a common codec such as ulaw on both legs.
4. Carrier side. If ports, IP, and codec all check out, the carrier may simply not be sending media yet, or may be holding it until it sees inbound RTP. Confirm the trunk is registered and that the carrier expects symmetric media.

Work the list top to bottom and you will land on the broken leg fast. For the wider method, see the troubleshooting playbook. When only one party hears audio, the cause is different, so read one-way audio and how to find the cause next. If you would rather not chase RTP and NAT on a self-managed box, VICIfast runs the server so the ports and IPs are right from the start.