When I read a file from Google Drive and pass it to Whisper, it attaches an extension .mpga instead of the original audio extension (mp3, m4a). So Openai gives an invalid file type error.
This was discussed earlier too and seems to be a bug.
Also, can any Activepieces piece compress the audio files? I am compressing the files externally as there is a 4 MB limit.
I too am having this issue. I have a mp3 file and testing with Activepieces turns it into a mpga file. I would like to know why!
When I manually download the file using the link in the Google Drive Read file piece, it’s a .mpga file. How strange, then, that my OpenAI Transcribe Audio piece complains that the file format is unsupported since it lists mpga as a supported format!
The problem with not being able to test this step is that Activepieces blocks me from completing my final Google Drive piece to create a new text file because I did not successfully run the OpenAI Transcribe Audio piece and don’t have its output to use.
Here is my flow:
Would really appreciate any help or workarounds. I’ve been unable to complete a single one of my Activepieces flows because of one issue or another like this.
I found a workaround to this on Discord.
See my example:
If the Transcribe Audio piece is, say, step_2 then you can bypass having to get test data from it first by using what seems to me to be a mustache template. In this example, I got the text element of the output with the notation: {{step_2.text}}
Like this:
I’ve now got my first working flow. Nice one, Activepieces team!
Many thanks for sharing! I will have to test this out.
