Skip to main content
Best answer

Extracting Twilio from authenticated Zendesk Ticket

  • November 6, 2025
  • 3 replies
  • 11 views

Forum|alt.badge.img

G’day All,

 

I’m using Zendesk and would like to extract customer voicemails for the purpose of AI Transcription (we’ve got a member of staff that is deaf, and transcriptions would permit them to be more collaborative)

Files appear to be hosted by Twilio (Phone Functionalist via Zendesk), and while it appears that the actual data doesn’t appear to be saved “within” the ticket, but does show a “recording_url” as a variable

File Source URL: ie.: https://myhelpdesk.zendesk.com/api/v2/channels/voice/calls/CA4fxxx3b38595xxxb323413508ff70691/twilio/voicemail/recording

 

What I can’t appear to do;

  1. Save the file as a .mp3 as I’m either having authentication issues (we don’t have a Twilio account, nor is the file “saved”/downloadable)
  2. Appear to use the Zendesk API to save the data (different session?).
  3. Find any variable from the “Find Latest Comment” in Zendesk that isn’t the “File Source URL” shown above.

 

Is there a simpler way to accomplish what I’m doing?

  • AI by Zapier won’t load anything without a URL (https://etc.com/audio.mp3)
  • Where I can access the file (Webhooks), I can’t save the audio contents to anything but a text file
  • Which leads to another list of problems.

Is there a way to “AI By Zapier” and pre-authenticate before running Gemini or any voice transcription AI Models?


Thank you for any insight here, its driving me a bit bonkers!

 

J.

This post has been edited by a moderator to remove personally identifiable information (PII). Please remember that this is a public forum and avoid sharing personal or potentially sensitive details.

Best answer by JAM-Ind

Hi Sam,

“I did it” - Thank you for your help!

The irony - CoPilot AI Addon for Zendesk is $50 per user license, however it will cost at least $2k for the license (we’ve got three seats, using two - If you want CoPilot AI The price is $50 per Seat, Per Month, no option for two licenses etc). Now I’ve got a more configurable product, and costing next to nothing.

 

For those of you possibly searching for a fix for Authentication (Webhooks) into Zendesk, see below

In the Basic Auth Field - Ensure you use the PIPE (|) symbol, the colon (:) referred to the  won't work

eg:bobby@zendesk.tables/token|MYFANCYAPIKEYFROMZENDESK

 

A Bug?

While I can use Webhooks to download the audio file (per your method), the file cannot be used in conjunction with “AI by Zapier” - whether its a temporary file, the file isn’t in the right format, or otherwise - the fix was to Upload the File using Google Drive, then Find the File before using that path for “AI by Zapier”

 

 

For those of you wanting to replace CoPilot, or just create voice transcripts;

AI Prompt:

Create a detailed transcription from the provided Audio File. Clearly identify each speaker in the conversation by labeling the first speaker as "Speaker #1" and the second speaker as "Speaker #2". If the person has provided their name or identified themselves, replace Speaker # with their name.

For accuracy and clarity, ensure that each new sentence begins with a timestamp indicating when it was spoken, formatted as [hh:mm:ss]. Additionally, indent the lines for Speaker #2 to enhance visual distinction. For example:

[00:00:01] Speaker #1:
Hello, how are you?

--- [00:00:03] Speaker #2:
I'm doing well, thanks for asking.

[00:00:01] Speaker #1:
How can I help you?

--- [00:00:01] Speaker #2:
I need to buy some lighting.

Results:

[00:00:00] Speaker #1:
Hello
--- [00:00:01] Speaker #2:
Hey, it's J from looking for Marina.
[00:00:03] Speaker #1:
Oh, she's not in today. Um, I think Marina was trying to confirm with you, um, that you got everything yesterday for reorder, we a plate.
--- [00:00:08] Speaker #2:
Yes.
[00:00:09] Speaker #1:
Yeah.
--- [00:00:11] Speaker #2:
I haven't time to get back to her, but you know.
[00:00:13] Speaker #1:
No, no, that's okay. Um, it's just clients are pushing us as you know, what it's like. So we're doing it back to you. Anyway, if there's anything else, um, should she just email it?
--- [00:00:20] Speaker #2:
Is that easier than calling?
[00:00:21] Speaker #1:
Oh, maybe. Yeah, I don't have time to sit and just answer.
--- [00:00:25] Speaker #2:
I know, I know. But if you email, I'll get that worked on then.
[00:00:27] Speaker #1:
You need a secretary.
--- [00:00:28] Speaker #2:
Yes. I had some of those. They weren't very good though.
[00:00:31] Speaker #1:
In a perfect world.
--- [00:00:32] Speaker #2:
Yeah.
[00:00:33] Speaker #1:
All right. Well, I'll let you go. You sound busy.
--- [00:00:35] Speaker #2:
I am. Thank you.
[00:00:35] Speaker #1:
Yeah, that's all right. No problem. Have a good weekend.
--- [00:00:38] Speaker #2:
Cool. Enjoy the weekend. Bye.
[00:00:39] Speaker #1:
Bye.

 

3 replies

SamB
Community Manager
Forum|alt.badge.img+11
  • Community Manager
  • 9676 replies
  • November 6, 2025

Hi and welcome to the Community ​@JAM-Ind! 🎉

Hmm, it looks like the Data Recording Url field might be providing an API endpoint, not a file URL which is strange. 🤔 I wonder if you could try using an API Request action to grab the file if you use the value in the Data Recording Url for the URL field and select GET for the HTTP Method field.

Can you give that a try and let me know if that works?


Forum|alt.badge.img
  • Author
  • New
  • 1 reply
  • Answer
  • November 7, 2025

Hi Sam,

“I did it” - Thank you for your help!

The irony - CoPilot AI Addon for Zendesk is $50 per user license, however it will cost at least $2k for the license (we’ve got three seats, using two - If you want CoPilot AI The price is $50 per Seat, Per Month, no option for two licenses etc). Now I’ve got a more configurable product, and costing next to nothing.

 

For those of you possibly searching for a fix for Authentication (Webhooks) into Zendesk, see below

In the Basic Auth Field - Ensure you use the PIPE (|) symbol, the colon (:) referred to the  won't work

eg:bobby@zendesk.tables/token|MYFANCYAPIKEYFROMZENDESK

 

A Bug?

While I can use Webhooks to download the audio file (per your method), the file cannot be used in conjunction with “AI by Zapier” - whether its a temporary file, the file isn’t in the right format, or otherwise - the fix was to Upload the File using Google Drive, then Find the File before using that path for “AI by Zapier”

 

 

For those of you wanting to replace CoPilot, or just create voice transcripts;

AI Prompt:

Create a detailed transcription from the provided Audio File. Clearly identify each speaker in the conversation by labeling the first speaker as "Speaker #1" and the second speaker as "Speaker #2". If the person has provided their name or identified themselves, replace Speaker # with their name.

For accuracy and clarity, ensure that each new sentence begins with a timestamp indicating when it was spoken, formatted as [hh:mm:ss]. Additionally, indent the lines for Speaker #2 to enhance visual distinction. For example:

[00:00:01] Speaker #1:
Hello, how are you?

--- [00:00:03] Speaker #2:
I'm doing well, thanks for asking.

[00:00:01] Speaker #1:
How can I help you?

--- [00:00:01] Speaker #2:
I need to buy some lighting.

Results:

[00:00:00] Speaker #1:
Hello
--- [00:00:01] Speaker #2:
Hey, it's J from looking for Marina.
[00:00:03] Speaker #1:
Oh, she's not in today. Um, I think Marina was trying to confirm with you, um, that you got everything yesterday for reorder, we a plate.
--- [00:00:08] Speaker #2:
Yes.
[00:00:09] Speaker #1:
Yeah.
--- [00:00:11] Speaker #2:
I haven't time to get back to her, but you know.
[00:00:13] Speaker #1:
No, no, that's okay. Um, it's just clients are pushing us as you know, what it's like. So we're doing it back to you. Anyway, if there's anything else, um, should she just email it?
--- [00:00:20] Speaker #2:
Is that easier than calling?
[00:00:21] Speaker #1:
Oh, maybe. Yeah, I don't have time to sit and just answer.
--- [00:00:25] Speaker #2:
I know, I know. But if you email, I'll get that worked on then.
[00:00:27] Speaker #1:
You need a secretary.
--- [00:00:28] Speaker #2:
Yes. I had some of those. They weren't very good though.
[00:00:31] Speaker #1:
In a perfect world.
--- [00:00:32] Speaker #2:
Yeah.
[00:00:33] Speaker #1:
All right. Well, I'll let you go. You sound busy.
--- [00:00:35] Speaker #2:
I am. Thank you.
[00:00:35] Speaker #1:
Yeah, that's all right. No problem. Have a good weekend.
--- [00:00:38] Speaker #2:
Cool. Enjoy the weekend. Bye.
[00:00:39] Speaker #1:
Bye.

 


SamB
Community Manager
Forum|alt.badge.img+11
  • Community Manager
  • 9676 replies
  • November 7, 2025

That’s awesome news ​@JAM-Ind - nice work! 🙌 I’m so glad you were able to get it sorted and really appreciate you following up here to share details of the solution you put in place. 🤗

I’m not entirely sure whether that’s caused by a bug, but if you contact our Support directly they’ll be able to take a look into the logs for your account to investigate that further. If it is a bug then they’ll be able to submit a bug report to our engineers to get it fixed.

It seems like you’re all good for the moment though, but feel free to reach back out if you need help with anything further - aways happy to lend a hand. 

In the meantime, happy Zapping! ⚡