How to Make Pipedream ChatGPT Whisper Transcription Accept a Specific Audio URL That Zapier Accepts?

Help
1

This topic was automatically generated from Slack. You can find the original thread here.

Greetz any clues how I can get the Pipedream ChatGPT Whisper Transcription to accept a audio URL like this: https://storage.tally.so/private/TBR_315_-_11_9_23_336_PM7s9n4.mp3?id=1XyPQl&accessToken=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpZCI6IjFYeVBRbCIsImZvcm1JZCI6Im1Sb044bCIsImlhdCI6MTcwNjQxODU3MH0.Hp3PXeMGfx855ed3HGlq6eGS414i-7kVfD0mYRWCvNY&signature=b8b1b743e9a77c93e5928ebd37531e406871d7f91bab1b90e707c7345a0650dd ?? The URL displays to me as a direct audio file (mp3) but Pipedream errors out with invalid file type. FYI Zap does accept it for the same type of use case. Hellllllo please :pleading_face:

2

Hi Gregory, thanks for the link. I can confirm this link isn’t in the proper format that you’ll need for Whisper.

The issue is that this is URL opens a webpage, not the actual file.

Whisper needs a link that accesses the MP3 directly, not a link that returns a webpage that can be used to play the mp3 as a user.

3

Thanks for the quick reply. Any suggestions for accessing the file via Pipedream for Whisper? ideally I’d like to get it to Whisper in such a manner that I get the benefit of chunking if needed to get past the 24 MB cap .

4

Hm weird, I just tried to download the file to /tmp and I uploaded it as a Project file too to make sure it was being downloaded correctly.

Yet Whisper is throwing the same error about invalid formatting.

5

It’s definitely a mp3 (right? I think)

6

I believe so, I’m seeing the same issue with this sample mp3:

7

You can report bug reports on actions and triggers here in our public Pipedream Github Repository:

A Pipedream team developer or community member will investigate and make a new version to attempt to fix it if they can reproduce the bug.

All Pipedream triggers and actions are open source, so you can also contribute your own fix if you’d like.

Learn about how to develop new and modify existing components here: Overview

8

Awesome, I would have found a bug in this awesome platform :grinning:

9

Any suggestions as to workaround?

10

You can use Node.js or Python code steps with your OpenAI API key instead if it is in fact a bug with that specific action.

11

It may be an issue on OpenAIs side, or perhaps a new requirement on that Whisper API route to pass an explicit file type argument.

12

To be clear, you suspect the “Send to Whisper for transcript” action is where the issue may lie, right?

13

I’ve been testing with the Whisper - Create Transcription action

14

Under the OpenAI app in Pipedream

15

Yes that’s the one I meant

16

Let me get to Python-ing :slight_smile: thx again