The call transcription feature automatically converts call recordings into text using AI. It automatically transcribes every new call recording in the background, and the finished transcript is displayed in the ticket right below the recording. The transcription is managed by OpenAI's Whisper speech-to-text model.
Note: To have access to the call transcription feature, it is required to have call recording enabled for your phone number or call button, and your account has to be integrated with an OpenAI AI provider.
How it works
As soon as a call ends and its recording is created, the system requests a transcription of it and the audio is sent to your OpenAI provider. While the AI is working on the transcription, the system displays the "Transcription in progress..." message. The message is automatically replaced with a button once the transcript is ready, it is not required to reload the ticket.
Once the transcription is generated, the AI provider returns the text together with the language it detected. The whole workflow is automatic, so agents are not required to initiate or manage the transcription process. Once finished, the transcript simply appears in the ticket.
If the transcript could not be generated, a short failure message is displayed instead. See Limits and unsuccessful transcriptions for more details.
Note: At this moment, transcription applies only to recordings created since the feature was enabled. Already existing recordings are not retroactively transcribed.
Enabling call transcription
To enable the feature, navigate to Configuration > Call > Settings and check the "Enable call transcription" checkbox, and save the settings.
Note: The checkbox is displayed only if an AI provider is connected in your account.
Tip: The audio is sent to your configured AI provider for processing, which means an external service processes the call audio. By enabling this feature, you confirm that recording and transcribing your calls complies with the laws applicable in your country and your customers' countries. The standard way to obtain consent is an IVR welcome message, played to the caller when the call is initiated, announcing that the call is being recorded and will be processed by a 3rd-party AI company.
Accessing a call transcript in a ticket
The generated transcript is located in the same place as the call recording to which it belongs, below the audio player with the given recording. Click the "Show transcript" button to display it. If the AI was able to identify the language of the call, the transcript displays the recognized language as well.
Limits and unsuccessful transcriptions
Not every recording can be turned into text. Recordings longer than 30 minutes are not transcribed, so very long calls are labeled with a failure message instead of a partial transcript.
When a transcript cannot be produced, the system shows one of the following reasons:
- "Recording too long to transcribe" - the recording exceeds the supported 30-minute length.
- "Transcription is temporarily unavailable. Please try again later." - your AI provider could not be reached, or was rate-limited.
- "The recording was deleted before it could be transcribed" - the given recording, or its whole ticket, was deleted while the transcription was still in process.
- "Transcription is not available for your account." - the AI provider rejected the request, most commonly because the connected OpenAI API key is invalid or lacks access to the Whisper model.
At this moment, a recording that failed to be transcribed is not picked up again automatically, and it cannot be retried. A call whose recording was not created (because call recording was not enabled for that phone number or call button) has no transcript section at all.