ChatGPT Transcription control

Questions
1

Describe the problem/error/question

Hi, I want to pass an audio recording for transcription. I have the process of picking up a file from Dropbox and transcribing it working, but I need to control the prompt passed to ChatGPT so it understands how to handle the audio file. I need to use the stereo file and have the conversation broken out into separate speakers (2 in total). I need to timestamp the conversation so I can build it into a chat flow so it can be visually followed.

Any guidance is appreciated.

What is the error message (if any)?

Please share your workflow

image
image2742×720 54 KB

(Select the nodes on your canvas and use the keyboard shortcuts CMD+C/CTRL+C and CMD+V/CTRL+V to copy and paste the workflow.)

Share the output returned by the last node

Information on your n8n setup

  • n8n version:
  • Hosted
  • Database (default: SQLite):
  • No DB
  • n8n EXECUTIONS_PROCESS setting (default: own, main):
  • Running n8n via (Docker, npm, n8n cloud, desktop app):
  • Operating system:
2

What you are looking for is called diarization and afaik OpenAI Whisper does not support diarization out of the box. I have used this model on replicate before for podcast transcription and speaker labelling: thomasmol/whisper-diarization – Run with an API on Replicate. You would have to connect it via HTTP, but it works pretty well. Hope this helps.

3

Much appreciated. I will take a look.

4

I am still investigating, but this seemed like a great suggestion.