Real-time audio streaming APIs for contact center AI

See also

Sources

1. 1
   Audio that Amazon Connect sends to Kinesis Video Streams uses an 8 kHz sampling rate.

   Plan for live media streaming from Connect Customer to Kinesis Video Streams · Plan for live media streaming, opening paragraphs · Checked 2026-10-02
2. 2
   The Amazon Connect Start and Stop media streaming blocks are each configured to specify what to stream: From the customer or To the customer.

   Example flow for testing live media streaming in Connect Customer · Procedure step 4 · Checked 2026-10-02
3. 3
   Amazon Connect can stream what the customer hears (agent speech and system prompts) and what the customer says (including while on hold), and the streams include Amazon Lex bot interactions in the flow.

   Set up live media streaming of customer audio in Connect Customer · Set up live media streaming of customer audio, bulleted list and following paragraph · Checked 2026-10-02
4. 4
   AWS points developers to the amazon-connect/connect-kvs-consumer-demo repository, which uses the Kinesis Video Streams Parser Library to read both tracks and store them as raw PCM.

   Develop live media streaming in Connect Customer · Develop live media streaming, paragraphs 1-2 · Checked 2026-10-02
5. 5
   The Amazon Connect Start media streaming block can be used in Inbound, Customer Queue, Agent Whisper, Customer Whisper, Outbound Whisper, Transfer to Agent and Transfer to Queue flows.

   Flow block in Connect Customer: Start media streaming · Flow types · Checked 2026-10-02
6. 6
   The KMS key chosen for Amazon Connect live media streaming must be in the same Region as the instance and be either a customer managed key or the aws/kinesisvideo AWS managed key.

   Enable live media streaming in your Connect Customer instance · Step: encryption, key restrictions 1-2 · Checked 2026-10-02
7. 7
   Amazon Connect live media streaming sends customer audio to an Amazon Kinesis video stream.

   Set up live media streaming of customer audio in Connect Customer · Set up live media streaming of customer audio, first paragraph · Checked 2026-10-02
8. 8
   If an Amazon Connect instance selects No data retention for live media streaming, the audio is available to consume for only 5 minutes, the Kinesis default minimum.

   Enable live media streaming in your Connect Customer instance · Step: Data retention period > Important · Checked 2026-10-02
9. 9
   Amazon Connect live media streaming is not enabled by default; it is turned on per instance under Data storage > Live media streaming.

   Enable live media streaming in your Connect Customer instance · Enable live media streaming, intro and steps 3-4 · Checked 2026-10-02
10. 10
    Amazon Connect uses one Kinesis video stream per active call and creates more streams as needed until the account reaches the Kinesis Video Streams Number of Streams quota.

    Plan for live media streaming from Connect Customer to Kinesis Video Streams · Do you need to increase your service quotas? · Checked 2026-10-02
11. 11
    Amazon Connect keeps its PutMedia requests within the Kinesis Video Streams 5 TPS quota, so customers do not need to request an increase for it.

    Plan for live media streaming from Connect Customer to Kinesis Video Streams · Do you need to increase your service quotas? > Tip · Checked 2026-10-02
12. 12
    Amazon Connect uses Kinesis Video Streams multi-track support so that what the customer says is on a separate track from what the customer hears.

    Plan for live media streaming from Connect Customer to Kinesis Video Streams · Plan for live media streaming, opening paragraphs · Checked 2026-10-02
13. 13
    Amazon Connect exposes the stream ARN, start and stop timestamps, and start fragment number of the customer audio as contact attributes under $.MediaStreams.Customer.Audio.

    Contact attributes for live media streaming in Kinesis Video Streams · Attribute list (StreamARN, StartTimestamp, StopTimestamp, StartFragmentNumber) · Checked 2026-10-02
14. 14
    Amazon Connect captures customer audio until a Stop media streaming block runs, even if the contact passes to another flow; a Stop media streaming block is required to stop it.

    Flow block in Connect Customer: Start media streaming · Configuration tips · Checked 2026-10-02
15. 15
    The Amazon Connect Kinesis video streams expose the audio as tracks named AUDIO_TO_CUSTOMER and AUDIO_FROM_CUSTOMER.

    Develop live media streaming in Connect Customer · Develop live media streaming, paragraph 2 · Checked 2026-10-02
16. 16
    The Amazon Connect Start media streaming block supports only the voice channel; chat, task and email contacts take the Error branch.

    Flow block in Connect Customer: Start media streaming · Supported channels table · Checked 2026-10-02
17. 17
    Only 5 AudioHook Monitor integrations can stream audio for an interaction at the same time, and this limit is not configurable.

    AudioHook Monitor overview · Limits section · Checked 2026-10-02
18. 18
    AudioHook Monitor sends the configured API key to the service provider in an X-API-KEY header and uses the client secret to sign connection requests.

    Configure and activate AudioHook Monitor in Genesys Cloud · Credentials tab > API Key; Client Secret · Checked 2026-10-02
19. 19
    The Genesys Architect Audio Monitoring action starts or stops AudioHook Monitor streaming to a third party.

    AudioHook Monitor overview · Overview, activation description · Checked 2026-10-02
20. 20
    Genesys charges for AudioHook Monitor while connected but not during secure pause or client-initiated pause.

    AudioHook Monitor overview · Overview, billing statement · Checked 2026-10-02
21. 21
    A Genesys Cloud organisation can have at most 20 AudioHook Monitor integrations.

    AudioHook Monitor overview · Limits section · Checked 2026-10-02
22. 22
    AudioHook Monitor is not supported on BYOC Premises or other premises-based Edge deployments.

    AudioHook Monitor overview · Limits / requirements section · Checked 2026-10-02
23. 23
    AudioHook Monitor is a Genesys Cloud premium application that must be installed from Genesys AppFoundry before it can be configured.

    Configure and activate AudioHook Monitor in Genesys Cloud · Prerequisites · Checked 2026-10-02
24. 24
    Genesys AudioHook Monitor streams voice interaction audio and metadata from Genesys Cloud to third-party service endpoints in near real time; it is a streaming mechanism, not storage.

    AudioHook Monitor overview · Overview, opening section · Checked 2026-10-02
25. 25
    Genesys secure pause automatically pauses AudioHook Monitor streaming during PCI-relevant flows, and the third-party service can also pause, resume and end the stream.

    AudioHook Monitor overview · Overview, pause and secure pause statements · Checked 2026-10-02
26. 26
    AudioHook Monitor's Channel setting is external (customer), internal (agent) or both; both produces a stereo stream with external on the left and internal on the right.

    Configure and activate AudioHook Monitor in Genesys Cloud · Properties tab > Channel · Checked 2026-10-02
27. 27
    The AudioHook Monitor Connection URI must begin with wss:// rather than https://.

    Configure and activate AudioHook Monitor in Genesys Cloud · Properties tab > Connection URI · Checked 2026-10-02
28. 28
    The Genesys AudioHook reference implementation handles 16-bit linear PCM and PCMU audio at 8000 Hz in mono and stereo.

    audiohook-reference-implementation · README, protocol and audio support section · Checked 2026-10-02
29. 29
    Genesys says its AudioHook reference implementation is a sample blueprint that lacks some robustness and resiliency patterns needed for a production service.

    audiohook-reference-implementation · README, disclaimer · Checked 2026-10-02
30. 30
    Third-party AI that consumes Twilio Media Streams or Amazon Connect live media streaming must accept 8 kHz narrowband audio or resample it, because neither documents a wideband option.inferred

    Media Streams - WebSocket Messages · start message mediaFormat (8000 Hz); compare AWS plan page 8 kHz statement · Checked 2026-10-02
31. 31
    Twilio Media Streams media messages carry the audio as a base64-encoded payload inside JSON text messages.

    Media Streams - WebSocket Messages · Messages from Twilio > media message · Checked 2026-10-02
32. 32
    A bidirectional Twilio Media Stream can receive only the inbound track.

    Media Streams Overview · Bidirectional Media Streams · Checked 2026-10-02
33. 33
    A Twilio call can have only one bidirectional Media Stream.

    Media Streams Overview · Bidirectional Media Streams, limits · Checked 2026-10-02
34. 34
    A bidirectional Twilio Media Stream lets the WebSocket application both receive call audio and send audio that Twilio plays on the call.

    Media Streams Overview · Bidirectional Media Streams · Checked 2026-10-02
35. 35
    A bidirectional Twilio Media Stream is started with <Connect><Stream>, which blocks subsequent TwiML instructions until the WebSocket connection is disconnected.

    Media Streams Overview · Bidirectional Media Streams · Checked 2026-10-02
36. 36
    Twilio unidirectional Media Streams do not support DTMF; bidirectional streams support DTMF only from Twilio to the media server.

    Media Streams Overview · Unidirectional and Bidirectional Media Streams sections · Checked 2026-10-02
37. 37
    An application on a bidirectional Twilio Media Stream can send media, mark and clear messages to Twilio.

    Media Streams - WebSocket Messages · WebSocket messages to Twilio · Checked 2026-10-02
38. 38
    Twilio sends connected, start, media, dtmf, stop and (bidirectional only) mark messages over a Media Streams WebSocket.

    Media Streams - WebSocket Messages · WebSocket messages from Twilio · Checked 2026-10-02
39. 39
    Twilio Media Streams carry audio as audio/x-mulaw at 8000 Hz, one channel.

    Media Streams - WebSocket Messages · Messages from Twilio > start message > mediaFormat; Send a media message · Checked 2026-10-02
40. 40
    Audio a bidirectional Twilio stream sends back must not include audio file header bytes, or the media is streamed incorrectly.

    Media Streams - WebSocket Messages · Messages to Twilio > media message · Checked 2026-10-02
41. 41
    A unidirectional Twilio Media Stream is started with the <Start><Stream> TwiML verb or through the Stream resource REST API.

    Media Streams Overview · Unidirectional Media Streams · Checked 2026-10-02
42. 42
    A unidirectional Twilio Media Stream can be stopped with <Stop><Stream name=...> or through the REST API.

    TwiML Voice: <Stream> · Stop a unidirectional stream · Checked 2026-10-02
43. 43
    The TwiML <Stream> track attribute accepts inbound_track (the default), outbound_track or both_tracks, and applies only to unidirectional streams.

    TwiML Voice: <Stream> · Attributes > track · Checked 2026-10-02
44. 44
    A Twilio call can carry up to four tracks simultaneously across unidirectional Media Streams, a limit shared with SIPREC, Real-Time Transcription and Answering Machine Detection.

    Media Streams Overview · Unidirectional Media Streams, limits · Checked 2026-10-02
45. 45
    A unidirectional Twilio Media Stream lets the WebSocket application receive call audio but not send audio back for playback on the call.

    Media Streams Overview · Unidirectional Media Streams · Checked 2026-10-02
46. 46
    The TwiML <Stream> url does not support query string parameters; custom key-value data is passed with nested <Parameter> elements instead.

    TwiML Voice: <Stream> · Attributes > url; Custom parameters · Checked 2026-10-02
47. 47
    Twilio Media Streams deliver live call audio over a WebSocket connection to a destination the customer specifies.

    Media Streams Overview · Media Streams overview, introduction and network requirements · Checked 2026-10-02
48. 48
    The url attribute of the TwiML <Stream> noun accepts only the wss protocol.

    TwiML Voice: <Stream> · Attributes > url · Checked 2026-10-02

Documents

Cite this page

WarmTransfer. (2026, October 2). Real-time audio streaming APIs for contact center AI. WarmTransfer. https://warmtransfer.net/knowledge/contact-center-audio-streaming-apis

@misc{warmtransfer-contact-center-audio-streaming-apis,
  title  = {Real-time audio streaming APIs for contact center AI},
  author = {{WarmTransfer}},
  year   = {2026},
  url    = {https://warmtransfer.net/knowledge/contact-center-audio-streaming-apis},
  note   = {Verified 2026-10-02}
}