Voicemail Transcription


Uses API services to transcribe voicemails into text to be used in the app-sms and the voicemail to email options. Bing’s Speech API or other generic APIs can be used.

Bing API

Sign up and language information is located on Microsoft Site Note: The Bing Speech API is deprecated as of October 2018, this works for now but needs to be ported to the new API

Warning

We cannot use mod_shout to record Voicemails because the transcription service needs an uncompressed version of the audio. Therefore we will record in WAV and then use LAME to re-encode in MP3. This could cause added resource utilization to your system.

Goto Advanced > Default Settings. Add the following entries

Category Subcategory Type Value Enabled
voicemail transcribe_provider text microsoft True
voicemail microsoft_key1 text {your microsoft key #1} True
voicemail microsoft_key2 text {your microsoft key #2} True
voicemail transcribe_language text en-US True
voicemail transcribe_enabled boolean true True

Click “Reload” at the top of the page.

Goto Status > Sip Status.

Click “Flush Memcache”, “Reload XML” and “Rescan”.

If you entered your key’s correctly, you should now start getting transcriptions delivered in your voicemail to email and you will also see them on the Messages page.

IBM Watson API

Sign up and language information is located on IBM Watson’s Site

Warning

We cannot use mod_shout to record Voicemails because the transcription service needs an uncompressed version of the audio. Therefore we will record in WAV and then use LAME to re-encode in MP3. This could cause added resource utilization to your system.

Goto Advanced > Default Settings. Add the following entries

Category Subcategory Type Value Enabled
voicemail transcribe_provider text watson True
voicemail watson_key text { your watson key } True
voicemail watson_url text { *watson url } True
voicemail transcribe_language text en-US True
voicemail transcribe_enabled boolean true True
voicemail json_enabled boolean true True

NOTE: Watson URL used for testing was the following: https://stream.watsonplatform.net/speech-to-text/api/v1/recognize?model=en-US_NarrowbandModel

Click “Reload” at the top of the page.

Goto Status > Sip Status.

Click “Flush Memcache”, “Reload XML” and “Rescan”.

If you entered your key’s correctly, you should now start getting transcriptions delivered in your voicemail to email and you will also see them on the Messages page.

Custom API

API info from the Speech to Text provider of your choice is needed, or you can self host a transcription engine like Mozilla DeepSpeech or Kaldi ASR

Goto Advanced > Default Settings. Add the following entries

Category Subcategory Type Value Enabled Required?
voicemail transcribe_provider text custom True  
voicemail transcription_server text https://yourserver True  
voicemail json_enabled boolean true True Optional
voicemail api_key text your_api_key True Optional
voicemail transcribe_language text en-US True  
voicemail transcribe_enabled boolean true True  

Click “Reload” at the top of the page.

Goto Status > Sip Status.

Click “Flush Memcache”, “Reload XML” and “Rescan”.

If you entered your key’s correctly, you should now start getting transcriptions delivered in your voicemail to email and you will also see them on the Messages page.