Programmable Voice, VoiceML

Move compatible Programmable Voice workloads without rewriting call logic.

VoiceML is VoiceTel's XML-based voice runtime for teams that already have voice markup, callbacks, recordings, conferences, queues, and API-driven call flows. Keep the interaction model; move the economics and support path to VoiceTel.

Build voice workflows on VoiceTel without paying retail markup on every layer of the call. Use API-driven calling, SIP access, recordings, conferences, streams, transcriptions, queues, and status callbacks on a platform built for telecom teams.

Start migration Compatibility matrix

VoiceML example

The markup stays readable, inspectable, and easy to test before cutover. A small inbound IVR — gather a digit, route to a queue, fall back to voicemail:

<?xml version="1.0" encoding="UTF-8"?>
<Response>
    <Say voice="Polly.Joanna">Welcome to Acme Support. Press 1 for billing, 2 for technical support.</Say>
    <Gather numDigits="1" action="/route" method="POST" timeout="6">
        <Say>You can also stay on the line to reach the next available agent.</Say>
    </Gather>
    <Dial timeout="20">
        <Queue>support</Queue>
    </Dial>
    <Record maxLength="120" transcribe="false" action="/voicemail" />
    <Say>Thanks. We'll follow up within one business day.</Say>
</Response>

Webhook signatures, parameter names, and verb behavior match Twilio's voice markup, so an existing handler at /route or /voicemail works unchanged once you point your application's host at voiceml.voicetel.com.

Compatible voice surface

  • Calls

    Create, fetch, update, list, and end calls using the supported voice-call lifecycle surface.

  • VoiceML call control

    Answer, route, record, gather, dial, queue, conference, redirect, stream, and transcribe calls using XML-based voice markup.

  • Recordings

    Start, pause, resume, stop, list, retrieve, and delete supported recordings.

  • Conferences and queues

    Use conference resources, participants, recordings, queue CRUD, and queue members where supported.

  • Streams, transcriptions, and SIP recording

    Start and inspect supported live media, transcription, and SIP recording resources with documented caveats.

  • AI voice agents

    Bridge live calls to your speech-to-text + LLM + text-to-speech pipeline with <Connect><ConversationRelay> — a Twilio-compatible WebSocket relay that handles barge-in, DTMF forwarding, and graceful handoff.

  • Applications

    Save voice-only application defaults and resolve them from call creation where supported.

  • Signed webhooks

    Status, recording, gather, dial, conference, AMD, and action callbacks use signed webhook delivery for request verification.

  • Free defaults, full vendor choice

    Free English speech-to-text and built-in text-to-speech ship with every account so you can prototype the same hour. For production, pick the STT, TTS, and storage vendor that fits your accuracy, latency, language, and data-handling needs — VoiceTel does not mark up third-party costs.

Detailed guides

  • Call recording and SIPREC

    Capture WAV files to your own Amazon S3 bucket via <Record> / <Dial> / <Conference>, or fork live media to your SIPREC-compatible Session Recording Server with <Start><Siprec>.

  • Speech recognition with <Gather>

    Free English default for quick prototyping, then switch to Deepgram, AWS Transcribe, Azure Speech, Google, OpenAI, or whisper.cpp when you're ready to tune for production. Same selection powers <Connect><ConversationRelay>.

  • Text-to-speech with <Say>

    Built-in voices in English, Spanish, French, and German, or bring your own Amazon Polly, Azure, ElevenLabs, Deepgram Aura, or Cartesia Sonic account for studio-quality speech.

What's different

  • Voice-first compatibility — The compatible surface centers on voice: Calls, Conferences, Queues, Applications, Recordings, IncomingPhoneNumbers, and the full SIP/Domains, SIP/CredentialLists, and SIP/IpAccessControlLists trunking surface (including domain ↔ list mappings and the v2/SipDomains Inbound Processing Region resource). Basic SMS send / list / fetch / delete is also available via /Messages when a per-tenant provider is configured. MMS, number provisioning, pricing APIs, and other CPaaS products use native VoiceTel product APIs or portal workflows.
  • Payment verb gated — The payment verb is fully compatible. Card capture is enabled per account — contact support to turn it on.
  • Number provisioning path — DIDs are managed through VoiceTel number workflows, not through the compatible customer-facing voice API surface.

VoiceML FAQ

Can existing Twilio voice applications migrate?

Yes, when they stay within the supported VoiceML-compatible voice surface. Review the compatibility matrix before production cutover.

Which text-to-speech providers does VoiceTel support?

Supported text-to-speech integrations. Open-source engines run on VoiceTel infrastructure; cloud vendors are billed by you directly.

Which speech-to-text providers does VoiceTel support?

English calls have a free VoiceTel STT default for quick starts. For production — or for any non-English call — pick the vendor that's right for your accuracy, latency, language, and data-handling needs, and pay them directly with no VoiceTel markup. Supported integrations:

How are rates displayed?

Current public rates are on the pricing page, with a calculator that applies volume-tier discounts. Your customer portal displays the same rates plus any account-specific contract pricing.

What's better

  • Line-item pricing. Every rate published on the pricing page. Speech, recording storage, and TTS billed by you directly to your chosen provider — VoiceTel does not mark up third parties.
  • Operating since 1999. 27 years in business — ISP origins, now full-stack carrier with engineers who read SIP traces, not qualification scripts.
  • No platform tax. Inbound messaging included. The published per-minute rate is what gets billed — no per-minute platform fee on top.
  • Honest compatibility. Every supported verb and attribute labelled in the compatibility matrix. No silent feature gaps mid-migration.