Transcription changes in v20

[Callinectes]

From what I can tell, with v20 it is now required to use Google storage for recording and voicemail archives in order to use transcription. Am I correct? If so what happens to systems that have been storing recordings in a different type of storage for archive? Will they no longer be able to use transcription? This could also prove to be much more expensive than alternative storage locations.

 

[Callinectes]

@illcsales Microsoft Co Pilot - Yes makes sense of course, especially now with the Google Storage requirement. But right now is a bad time for new features so we will need some months.

It has been some months. Is there any chance there is a migration option, at least in planning or development, for transitioning from v18 with FTP storage for recording archives and transcribing voicemail messages to v20 with the same features? Even an option to move the recordings would help. Right now there is no migration path available to maintain this combined functionality, partly because of the Google Storage requirement for using any transcription features.

 

[ivank]

In SP2 you'll have the option to use alternative speech engine which will allow you to store your recordings anywhere and still have transcription.

 

[Callinectes]

In SP2 you'll have the option to use alternative speech engine which will allow you to store your recordings anywhere and still have transcription.

That's great to hear!

Is there an expected release date for SP2?

 

[ivank]

Not yet

 

[C.C.I]

There's some blatant misinformation going on here. This has nothing to do with the v1 or v2 Speech-to-Text API, and even if that were the case Google has made no mention of retiring the v1 API.

The limitation is that to transcribe audio longer than 60 seconds, the audio must be saved data to a Google Cloud Storage bucket and processed asynchronously. For short audio, 60 seconds or less, local audio files can still be provided and are processed synchronously: https://cloud.google.com/speech-to-text/docs/sync-recognize. This applies to both the v1 and v2 API: https://cloud.google.com/speech-to-text/v2/docs/sync-recognize "Audio content can be sent directly to Speech-to-Text from a local file, or Speech-to-Text can process audio content stored in a Cloud Storage bucket."

So this is down to 3CX choosing to switch all transcribing to the asynchronous api for audio processing, which requires files stored in a Cloud Storage bucket. I would love to know the percentage of voicemails (not recordings) that actually breach the 60 second mark.

I think a lot of the frustration comes from the lack of notification or guidance. In the v20 upgrade checklist page: https://www.3cx.com/blog/releases/v20-upgrade-checklist-faq/, it is targeted at preparing for v20, not what to do post upgrade. Yet some of the items (Step 2 and 3, regarding departments) are not possible pre-upgrade because Departments is not a feature of v18. And the only mention of Transcription is in the Step 9 FAQ in regards to how long v18 will remain available, but no instruction on steps that will need to be taken for transcription to continue working. 3CX simply broke the feature with no heads up.

The upgrade document should be split into a Pre-Upgrade and Post-Upgrade checklist, with callouts for breaking changes. Here's an example I just dealt where the vendor provided a great checklist document: https://www.visualsvn.com/support/topic/00233/

 

[NicholasP_3CX]

From Update 2 of Version 20 you can do the transcription via OpenAI https://www.3cx.com/docs/manual/openai-transcription/ independent of the storage location.

 

[C.C.I]

It would be nice if the documents would indicate what version things applied to. Some customers still have V18, but nowhere in the document does it even indicate this is related to V20. In this case not only do you need V20 but you also need Update 2, again, not mentioned anywhere.

We are on V20 Update 2 but there is no "Transcriptions" menu. The "OpenAI Whisper" menu is a direct child in the Integrations menu.