VonyX.HD
HD audio for AI agents, from any phone call
VonyX.HD turns the raw audio of a phone call into HD before it reaches your AI agent. VonyX speech models run in the call path and stream the result in real time, so your speech-to-text hears callers clearly, even on ordinary mobile calls.
Raw call. The phone network keeps almost nothing above 3.4 kHz, and the line adds noise.
What your agent hears today
Most phone calls still reach an AI agent as narrowband audio: sampled at 8 kHz, with almost nothing above about 3.4 kHz. That cuts much of the sound of letters like s, f and th, which is exactly where names, email addresses and account numbers go wrong. Add traffic, a speakerphone or a weak signal, and the agent is guessing.
Your agent hears the caller, not the line
- Clearer names, numbers and email addresses, the details that decide whether a call succeeds.
- Cleaner transcripts from the speech-to-text you already use.
- Fewer repeated questions and fewer wrong answers.
- Nothing to retrain. Same agent, same prompts, better input.
Same call. Two versions.
Raw call. The phone network keeps almost nothing above 3.4 kHz, and the line adds noise.
Works with the stack you run
Plans
| Feature | VonyX.HD | VonyX Standard |
|---|---|---|
| Local and toll-free numbers in 100+ countries | Yes | Yes |
| SIP over UDP or TCP, G.711 | Yes | Yes |
| Call logs and recordings | Yes | Yes |
| HD audio from VonyX speech models | Yes | No |
| WebSocket streaming as 16-bit PCM at 16 kHz | Yes | No |
| Direct ElevenLabs connection and webhook call status | Yes | No |
VonyX.HD questions
Still have a question?
No. Carrier HD Voice works only when both ends of a call support a wideband codec. VonyX.HD makes HD audio for your AI agent from ordinary narrowband calls, using VonyX speech models in the call path.
No. VonyX.HD improves the audio before it reaches your agent, so your speech-to-text, model and prompts stay the same.