I recently parted ways with YouTube TV due to its price increase and my family’s lack of television use. One thing that my wife missed was watching Friends. Friends were recently put behind a paywall on services such as MAX, so I decided to purchase the complete series collection on DVD—less than the price of one month of YouTube TV.
I had tried Plex before but lost interest. One reason was that it required me to ensure that my files were explicitly named and stored in concrete folder structures, which was time-consuming. The Friends collection contains 32 DVDs with 280+ episodes, so it should not be taken lightly. I could load them up on the server, but that’s not how I function. 🙂 I had my brand new $50 MacBook Air as a media server, so I was set.
One of the challenges is that there is much content besides just the episodes on each DVD. There are these strange multi-episode titles and a lot of bonus content that’s not of much interest. There is no mention of exactly what episode it is when you download it, and the episodes don’t have visual titles. I looked for titles around 23 minutes long, which worked well for digitizing the disk. (By the way, this is legal as I have the license to watch the DVD; I am making a secure copy for my use at home and not sharing it more broadly. I will also maintain the DVD – not turn around and sell it once I digitize it – to keep that license).
ChatGPT Showed me the proper file naming convention so that Plex would automatically add the thumbnails and metadata to my database.
My first approach to identifying which episode was which was to find a random guest star and use ChatGPT to find the episode information.
Once I gave it the command, using ChatGPT’s chat persistence, I could enter the person’s name and get a result. However, I did have to tell it to stop giving me additional information, which I’ve found to be standard.
This worked well, but it was super time-consuming. I had to open the video, scroll to the end, and then find a lower guest star (like Tom Selleck) to prevent them from appearing in more than one episode in that season.
Using my post-production background, I knew that each episode had a specific episode number at the end of the episode. I tried using that, and it worked, but it was just as time-consuming.
Then I realized that every time I put a DVD into my Mac Mini, it automatically started playing. I took a picture of the episode list from the menu (you can’t take screenshots of any DVD content on a Mac for security reasons) and then took a picture of the files digitized in their folder. I then told ChatGPT to write me an AppleScript to rename the files in the order in the folder in “Plex Format” using the image of the DVD menu as a guide. It took a few tries to tweak the AppleScript (I made a dupe of the files just in case – which I highly recommend when writing any code, AI or otherwise, to work on files).
Conclusion
This task demonstrates some basic principles of working with AI:
- If it’s boring or repetitive, let the computer do the work
- If your tool has access to the internet, doing “searches” can be baked into your prompt. I didn’t have to tell it to go out on the web to find the information – although I could have said it to look at a specific source
- Check everything. I found that sometimes it gave me the wrong answer for the guest star and the episode number. This is a great example here because a mistake would have only been a minor annoyance for my wife – which would eventually be a significant problem for me 🙂
- The more you know, the better AI works. From my experience, I knew about the production number. Given the way production works, I also learned to ask for a relatively minor guest star because they would most likely be a one-and-done-er. I also knew that I could use AppleScript to rename files and that ChatGPT does a good job of “reading” images.
- Don’t give up, and continually work to refine your process. If something is taking a lot of effort, think about a better way to do it and ask your AI bot for help.
