Messing around with OpenAI - API and Whisper

open-ai-api.md

Messing around with OpenAI

Disclaimer: this was just for fun and to learn about openai

I started messing around with openai's API and wanted to share some CLI tools. The following was used to test these scripts. May create a small container or codespace for this in the future.

Enviroment Used:

Fedora Silverblue 37 (VM):

• 4 x 2 Vcpus (8 cores total)

Packages Needed:

• ffmpeg (audio/video decoding tool)
• yt-dlp (download youtube vids fast)
• python3, pip, and Whisper (see Openai whisper git repo)
  • these packages can be installed with your pkg manager.

Creating Enviroment to Work in.

$ mkdir dev; cd dev; mkdir Text Audio

Convert Audio to mp3 with yt-dlp and useing Whisper to convert mp3 to English text.

$ cat <<EOF > commands.sh
#!/bin/bash
#Get URL from user.
read -p 'Video link: ' YOUTUBE_VIDEO_URL 

#Convert given URL to a .mp3.
yt-dlp -x --audio-format mp3 $YOUTUBE_VIDEO_URL \
	-o ~/dev/Audio/audio.mp3 

# Transcribe .mp3 to English Text.
whisper ~/dev/Audio/audio.mp3 --language English \ 
	--task transcribe --model base -o ~/dev/Text 
EOF

We now have succesfully transcribed a video and have text. Now what?

Check out whisper repo for more usage examples!

• https://github.com/openai/whisper

---

This Section is a work in progress

Add your API key to a .py file that is imported.

$ cat <<EOF > apiKey.py apiKeyOAI = '$YOUR KEY HERE' EOF

ToDo:

Cleanup .py file with API call to Completion endpoint.

quick notes on some neat things you can do with whisper. Still need to clean this up in future.