<Record> verb records the caller's voice and returns to you the URL of a file containing the audio recording. You can optionally generate text transcriptions of recorded calls by setting the 'transcribe' attribute of the
<Record> verb to 'true'.
If you choose to record voice or video calls, you need to comply with certain laws and regulations, including those regarding obtaining consent to record (such as California’s Invasion of Privacy Act and similar laws in other jurisdictions). Additional information on the legal implications of call recording can be found here.
Notice: Twilio recommends that you consult with your legal counsel to make sure that you are complying with all applicable laws in connection with communications you record or store using Twilio.
<Record> verb supports the following attributes that modify its behavior:
|Attribute Name||Allowed Values||Default Value|
|action||relative or absolute URL||current document URL|
|maxLength||integer greater than 1||
|recordingStatusCallback||relative or absolute URL||none|
|transcribeCallback||relative or absolute URL||none|
Use one or more of these attributes in a
<Record> verb like so:
The 'action' attribute takes a relative or absolute URL as a value. When recording is
finished Twilio will make a GET or POST request to this URL including the parameters
below. If no 'action' is provided,
<Record> will default to requesting the current
After making this request, Twilio will continue the current call using the
TwiML received in your response. Keep in mind that by default Twilio will
re-request the current document's URL, which can lead to unwanted looping
behavior if you're not careful. Any TwiML verbs occurring after a
There is one exception: if Twilio receives an empty recording, it will not make a request to the 'action' URL. The current call flow will continue with the next verb in the current TwiML document.
Twilio will pass the following parameters in addition to the standard TwiML Voice request parameters with its request to the 'action' URL:
|RecordingUrl||The URL of the recorded audio. The recording file may not yet be accessible when the 'action' callback is sent. Use recordingStatusCallback for reliable notification on when the recording is available for access.|
|RecordingDuration||The duration of the recorded audio (in seconds). To get a final accurate recording duration after any trimming of silence, use recordingStatusCallback.|
|Digits||The key (if any) pressed to end the recording or 'hangup' if the caller hung up|
A request to the
RecordingUrl will return a recording in binary WAV audio
format by default. To request the recording in MP3 format, append ".mp3" to the
The 'method' attribute takes the value 'GET' or 'POST'. This tells Twilio whether to request the 'action' URL via HTTP GET or POST. This attribute is modeled after the HTML form 'method' attribute. 'POST' is the default value.
The 'timeout' attribute tells Twilio to end the recording after a number of seconds of silence has passed. To disable this feature, set 'timeout' to 0. The default is 5 seconds.
The 'finishOnKey' attribute lets you choose a set of digits that end the recording when entered. For example, if you set 'finishOnKey' to '#' and the caller presses '#', Twilio will immediately stop recording and submit 'RecordingUrl', 'RecordingDuration', and the '#' as parameters in a request to the 'action' URL. The allowed values are the digits 0-9, '#' and '
*'. The default is '1234567890
*#' (i.e. any key will end the recording). Unlike
<Gather>, you may specify more than one character as a 'finishOnKey' value.
In some instances, when using the 'finishOnKey' attribute, the last part of the audio (approximately the one second before the key is pressed) may not be recorded.
The 'maxLength' attribute lets you set the maximum length for the recording in seconds. If you set 'maxLength' to '30', the recording will automatically end after 30 seconds of recorded time has elapsed. This defaults to 3600 seconds (one hour) for a normal recording and 120 seconds (two minutes) for a transcribed recording. Twilio Client calls using
<Record> are limited to 600 seconds (ten minutes).
The 'playBeep' attribute allows you to toggle between playing a sound before the start of a recording. If you set the value to 'false', no beep sound will be played.
The 'trim' attribute lets you specify whether to trim leading and trailing
silence from your audio files. 'trim' defaults to
trim-silence, which removes
any silence at the beginning or end of your recording. This may cause the
duration of the recording to be slightly less than the duration of the call.
The 'recordingStatusCallback' attribute takes a relative or absolute URL as an argument. If a 'recordingStatusCallback' URL is given, Twilio will make a GET or POST request to the specified URL when the recording is available to access.
Twilio will pass the following parameters with its request to the 'recordingStatusCallback' URL:
|AccountSid||The unique identifier of the Account responsible for this recording.|
|CallSid||A unique identifier for the call associated with the recording.|
|RecordingSid||The unique identifier for the recording.|
|RecordingUrl||The URL of the recorded audio.|
|RecordingStatus||The status of the recording. Possible values are:
|RecordingDuration||The length of the recording, in seconds.|
|RecordingChannels||The number of channels in the final recording file as an integer. Only
|RecordingSource||The initiation method used to create this recording.
This attribute indicates which HTTP method to use when requesting 'recordingStatusCallback'. It defaults to 'POST'.
recordingStatusCallbackEvent allows you to specify which recording status changes should generate a webhook to the URL specified in the recordingStatusCallback attribute. The available values are:
in-progress: the recording has started
completed: the recording is complete and available
absent:the recording is absent and inaccessible
To specify more than one value, separate each with a space. Default value is:
The 'transcribe' attribute tells Twilio that you would like a text representation of the audio of the recording. Twilio will pass this recording to our speech-to-text engine and attempt to convert the audio to human readable text. The 'transcribe' option is off by default. If you do not wish to perform transcription, simply do not include the transcribe attribute.
Note: transcription is a paid feature. If you include a
transcribeCallback attribute on your
Record verb, your account will be charged. See the pricing page for our transcription prices.
Additionally, transcription is currently limited to recordings initiated via the TwiML <Record> verb with a duration greater than 2 seconds and less than 120 seconds. If you request a transcription for a recording outside these duration limits, Twilio will write a warning to your debug log rather than transcribe the recording.
The 'transcribeCallback' attribute is used in conjunction with the 'transcribe' attribute. It allows you to specify a URL to which Twilio will make an asynchronous POST request when the transcription is complete. This is not a request for TwiML and the response will not change call flow, but the request will contain the standard TwiML request parameters as well as transcription specific ones.
If 'transcribeCallback' is specified, then there is no need to specify 'transcribe=true'. It is implied. If you specify 'transcribe=true' without a 'transcribeCallback', the completed transcription will be stored for you to retrieve later (see the REST API Transcriptions section), but Twilio will not asynchronously notify your application.
Twilio will not make a request to the provided 'transcribeCallback' URL if the voice recording is missing or absent.
The transcribeCallback URL attribute is considered Not PII.
Twilio will pass the following parameters with its request to the 'transcribeCallback' URL:
|TranscriptionSid||The unique 34 character ID of the transcription.|
|TranscriptionText||Contains the text of the transcription.|
|TranscriptionStatus||The status of the transcription attempt: either 'completed' or 'failed'.|
|TranscriptionUrl||The URL for the transcription's REST API resource.|
|RecordingSid||The unique 34 character ID of the recording from which the transcription was generated.|
|RecordingUrl||The URL for the transcription's source recording resource.|
|CallSid||A unique identifier for this call, generated by Twilio.|
|AccountSid||Your Twilio account id. It is 34 characters long, and always starts with the letters AC.|
|From||The phone number or client identifier of the party that initiated the call. Phone numbers are formatted with a '+' and country code, e.g.
|To||The phone number or client identifier of the called party. Phone numbers are formatted with a '+' and country code, e.g.
|CallStatus||A descriptive status for the call. The value is one of
|ApiVersion||The version of the Twilio API used to handle this call. For incoming calls, this is determined by the API version set on the called number. For outgoing calls, this is the API version used by the outgoing call's REST API request.|
|Direction||A string describing the direction of the call.
|ForwardedFrom||This parameter is set only when Twilio receives a forwarded call, but its value depends on the caller's carrier including information when forwarding. Not all carriers support passing this information.|
You can't nest any verbs within
<Record> and you can't nest
<Record> within any other verbs.
Twilio will execute the
<Record> verb causing the caller to hear a beep and the recording to start. If the caller is silent for more than 5 seconds, hits the '#' key, or the recording maxlength time is hit, Twilio will make an HTTP POST request to the default 'action' (the current document URL) with the parameters 'RecordingUrl' and 'RecordingDuration'.
This is example shows a simple voicemail prompt. The caller is asked to leave a message at the beep. The
<Record> verb beeps and begins recording up to 20 seconds of audio.
- If the caller does not speak at all, the
<Record>verb exits after 5 seconds of silence, falling through to the next verb in the document. In this case, it would fall through to the
- If the caller speaks for less than 20 seconds and is then silent for 5 seconds, Twilio makes a GET request to the 'action' URL. The
<Say>verb is never reached.
- If the caller speaks for the full 20 seconds, Twilio makes a GET request to the 'action' URL. The
<Say>verb is never reached.
Twilio will record the caller. When the recording is complete, Twilio will transcribe the recording and make an HTTP POST request to the 'transcribeCallback' URL with a parameter containing a transcription of the recording.
- Twilio will trim leading and trailing silence from your audio files. This may cause the duration of the files to be slightly smaller than the time a caller spends recording them.