Issue with Speech to Open AI to Notion Workflow Consuming Credits: Is Premium Worth It?

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

Hi, I hope I am writing the message in the right place. I have a Speech to Open AI to Notion workflow that keeps on just eating credits but not working at all. I am thinking of buying the premium but first I need to make sure that I am doing it all right otherwise it’s quite useless anyway to pay for it.
Is it ok if I describe the problem here?

Basically I have often very long meetings (min 15 to max 45 min each), various time per day and very often I need to write a breif to Notion. Now, the Speech to Open AI to Notion workflow would be exactly the solution I need but it doesn’t always work. It just eats credits, without working.

Usually it says that I am Out of Memory so I have increased to 1024 mb Memory and Timeout 300.

Once done that, it still doesn’t work and now it gives me this error 400 from Open AI.

at null.createError (/tmp/__pdg__/dist/code/8acd09b5bb3ac1f9052c9891ca552ef1c706b9f3fa6c3f646ebfb5cd5828f521/node_modules/.pnpm/axios@0.26.1/node_modules/axios/lib/core/createError.js:16:15)
    at null.settle (/tmp/__pdg__/dist/code/8acd09b5bb3ac1f9052c9891ca552ef1c706b9f3fa6c3f646ebfb5cd5828f521/node_modules/.pnpm/axios@0.26.1/node_modules/axios/lib/core/settle.js:17:12)
    at IncomingMessage.handleStreamEnd (/tmp/__pdg__/dist/code/8acd09b5bb3ac1f9052c9891ca552ef1c706b9f3fa6c3f646ebfb5cd5828f521/node_modules/.pnpm/axios@0.26.1/node_modules/axios/lib/adapters/http.js:322:11)
    at IncomingMessage.emit (events.js:412:35)
    at null.endReadableNT (internal/streams/readable.js:1333:12)
    at process.processTicksAndRejections (internal/process/task_queues.js:82:21)

Hello , in order for me to reproduce your issue, could you share the screenshot of your Open AI failed action?

I don’t know if this helps, I was just reading OpenAI docs and I though I read something about an input limit on requests. This was for chat API though, so I’m not sure if speech to text is different.

Hi , thanks so much for your prompt reply.
First of all I must say that I am no dev, just a product manager. But I have been following this tutorial to reach the outcome I wanted and it was at first working fine How to Take Perfect Notes with Your Voice Using ChatGPT and Notion

At the beginning I have followed to no-code version, then the hard-coded one because of the various issues related to memory and input limits.

But now, this is what comes out of the 5th step, the openai_chat one:

This is the step of the workflow tut I am referring to (and the code block I have pasted in that step) How to Take Perfect Notes with Your Voice Using ChatGPT and Notion

Hi , I’m so sorry for the late response. Would you mind expanding your Open AI Chat Code & Configuration section as image below?

Also, could you try the action with the shorter 5 minutes audio file to see if it works?

I suspect that the issue is because of the transcribe/audio length

Hi - I am very sorry for the laste reply but I was super busy with work yesterday. That is exactly why I need to workflow to work.

So here is what you asked

Thanks , could you try the action with the shorter 5 minutes audio file to see if it works?

Sorry for patching it so badly by the way!
Regarding uploading any other file, smaller, etc, nothing is working anymore.
But also, I have gone throught the entire hard coded workflow to bypass the limits of the file length. In the tutorial Frank says he managed to upload up to 90 minutes podcast. I just need to upload up to max 45 minutes meetings

If all this is simply for 5 minutes meetings then I’d rather do it manually. I really hope we can get out of this issue. Let me know please if it also could help if I pay

could you try the action with the shorter 5 minutes audio file to see if it works?
Hi , I asked you this to confirm if the issue is because of the audio length or not.

Also, could you tell me if your OpenAI account still have credit available so that you can run the transcribe?

I do the 5 minutes audio right now to check again anyway and send you the results

Open AI credit is good

Basically now it’s not even picking up the first trigger event