Capture call audio to your own storage or your own recorder.

Two recording paths, fully independent

Both methods are bring your own. VoiceTel does not store recordings on shared platform storage. Until you configure a destination, calls are not recorded and nothing is kept on our side.

Amazon S3 recordings

Store call recordings in your own AWS account. VoiceML uploads each recording to your S3 bucket using credentials you provide. You pay AWS directly — there is no VoiceTel markup on storage.

Enable S3 recording

Configure your destination from the Account page in the VoiceML portal (section Amazon recordings):

1. Open Account and scroll to Amazon recordings.
2. Enter your bucket details and IAM credentials (see table below).
3. Save. New recordings upload to your bucket on the next recorded call.

The secret access key is write-only in the portal: leave it blank to keep the existing value, or paste a new value to replace it.

Object key layout

Each recording is stored as:

<prefix>/<AccountSid>/<RecordingSid>.wav

For example, with prefix prod/recordings:

prod/recordings/AC<your-account-sid>/RE<recording-sid>.wav

AWS setup checklist

Before saving your portal settings:

1. Create an S3 bucket in your AWS account (private — block public access).
2. Create an IAM user scoped to that bucket only.
3. Attach a minimum policy allowing upload, download, and delete on objects in the bucket:

{
  "Version": "2012-10-17",
  "Statement": [{
    "Sid": "VoiceMLRecordings",
    "Effect": "Allow",
    "Action": ["s3:PutObject", "s3:GetObject", "s3:DeleteObject"],
    "Resource": "arn:aws:s3:::your-bucket-name/*"
  }]
}

4. Generate an access key for that IAM user and enter it in the portal.

Optional: enable bucket encryption (SSE-S3 or SSE-KMS) and a lifecycle policy to expire old recordings after your retention period.

When S3 recording applies

S3 recording is used by:

• The <Record> verb
• <Dial record="record-from-answer"> (and other non-do-not-record values)
• <Conference record="record-from-start"> (and similar)
• POST /Calls/{CallSid}/Recordings on the REST API

Using <Record>

The <Record> verb captures audio from the live call into a WAV file and uploads it to your S3 bucket.

Quick start

<?xml version="1.0" encoding="UTF-8"?>
<Response>
    <Record action="/handle-recording" method="POST"
            maxLength="120" playBeep="true" />
</Response>

When recording finishes, VoiceML posts to your action URL with Twilio-compatible form fields:

Common attributes

Voicemail-style flow

When action is omitted, VoiceML re-fetches the current document after the recording finishes:

<Response>
    <Say>Please leave a message after the tone.</Say>
    <Record maxLength="120" transcribe="true"
            transcribeCallback="/voicemail-transcript" />
    <Say>We did not receive a recording. Goodbye.</Say>
</Response>

Recording bridged and conference calls

<Dial>

Record the bridged portion of an outbound call:

<Response>
    <Dial record="record-from-answer"
          recordingStatusCallback="https://app.example/recording-status">
        <Number>+18005551234</Number>
    </Dial>
</Response>

Use recordingTrack="inbound", outbound, or both to select which leg is captured.

<Conference>

Record a conference room:

<Response>
    <Dial>
        <Conference record="record-from-start"
                    recordingStatusCallback="https://app.example/conf-recording">
            Support Queue
        </Conference>
    </Dial>
</Response>

S3 must be configured on your account for <Dial> and <Conference> recording to produce files.

Fetch and manage recordings

Download a recording

After a recording completes, fetch the WAV from the URL in RecordingUrl (or via the REST API):

GET /2010-04-01/Accounts/{AccountSid}/Recordings/{RecordingSid}.wav

Authenticate with your account SID and API key. The response redirects to a short-lived download URL for the object in your S3 bucket.

REST API

Recording metadata includes status, duration, date_created, and the public uri for the resource.

SIPREC recording

SIPREC (RFC 7866) sends a parallel copy of the call's media to your own Session Recording Server while the call continues normally. Use SIPREC when you already operate an enterprise recorder (compliance archive, quality monitoring, WORM storage, etc.) and want VoiceML to act as the Session Recording Client (SRC).

SIPREC is independent of Amazon S3 recording:

• S3 captures WAV files through VoiceML verbs.
• SIPREC streams live media to your SRS in real time.
• You can run both on the same call if your compliance design requires it.

Configure SIPREC connectors

A connector gives your SRS a name that VoiceML and the REST API can reference. Each connector points at one SRS SIP URI.

Portal setup

1. Open Account in the VoiceML portal.
2. In SIPREC recording, click Manage SIPREC connectors.
3. Click Add a connector and fill in:

4. Save. The connector is available on the next call.

SRS SIP URI format: use a standard SIP or SIPS URI with the host, port, and transport your recorder expects. Examples:

sip:10.0.1.50:5060
sip:recorder.example.com:5070;transport=tcp
sips:recorder.example.com:5061;transport=tls

Obtain the exact URI from your recorder vendor's documentation (see compatible Session Recording Servers below). Your SRS must be reachable from VoiceTel over the network paths your deployment uses.

Ensure your SRS firewall and security groups allow inbound SIP (and RTP) from VoiceTel's media and signaling addresses.

Using <Start><Siprec>

Start SIPREC on a live call, then continue with other voice-markup verbs:

<?xml version="1.0" encoding="UTF-8"?>
<Response>
    <Start>
        <Siprec connectorName="primary-srs"
                name="compliance-trail"
                track="both_tracks"
                statusCallback="https://app.example/siprec-status"
                statusCallbackMethod="POST">
            <Parameter name="tenant" value="acme-prod"/>
            <Parameter name="department" value="support"/>
        </Siprec>
    </Start>
    <Dial><Number>+18005551234</Number></Dial>
    <Stop><Siprec/></Stop>
</Response>

Attributes

<Parameter name="…" value="…"/> children pass metadata through to the SIPREC session. They appear on status callbacks as Parameter.<name> form fields.

Behavior notes

• One active SIPREC session per call. A second <Start><Siprec> replaces the first.
• <Stop><Siprec/> ends the fork; otherwise it ends when the call hangs up.
• connectorName is required. An empty or unknown name fails without starting a recording session.

REST API

Start or stop SIPREC on a live call without voice markup:

# Start (connector must exist and be enabled)
curl -X POST "https://voiceml.voicetel.com/2010-04-01/Accounts/AC…/Calls/CA…/Siprec" \
  -u "AC…:your_api_key" \
  --data-urlencode "ConnectorName=primary-srs"

# Stop
curl -X POST "https://voiceml.voicetel.com/2010-04-01/Accounts/AC…/Calls/CA…/Siprec/SR…" \
  -u "AC…:your_api_key" \
  --data-urlencode "Status=stopped"

Compatible Session Recording Servers

VoiceML implements SIPREC as a Session Recording Client and sends recording sessions to your SRS. Any recorder that accepts standard SIPREC INVITEs can work, subject to your network and codec requirements.

The table below lists widely deployed platforms customers connect to VoiceML. For each one, configure the recorder first, obtain its SRS SIP URI, then enter that URI in your VoiceML connector.

Example connector entries

After your vendor provides the SRS endpoint, register it in the portal:

Replace host, port, and transport with values from your vendor's deployment guide. NICE, Verint, and similar enterprise platforms often require TLS (sips:), specific ports, and allowlisting of VoiceTel's signaling addresses — follow the vendor checklist before going live.

Metadata parameters

Many enterprise recorders expect custom SIPREC metadata (tenant ID, agent ID, call type). Pass these as <Parameter> children on <Siprec>:

<Siprec connectorName="nice-compliance">
    <Parameter name="siteId" value="NYC-01"/>
    <Parameter name="recordType" value="compliance"/>
</Siprec>

Consult your recorder vendor for supported parameter names.

Migrating from Twilio

VoiceML uses the same voice markup and REST shapes Twilio documents:

On Twilio, recordings and SIPREC often run on Twilio-managed storage or connectors. On VoiceML, you configure your own S3 bucket and your own SRS endpoints — billing and data residency follow your AWS and recorder agreements.

Troubleshooting

Amazon S3 recordings

SIPREC

General