Advanced Voice Notes for Journaling On The Go

Saturday, October 26th, 2024, 8:33 AM

Using the method I will describe in this blog I wrote this entire blog with no edits in 3 minutes while walking the dogs. (Note: Link and Info Graphics were added later)

I do a lot of journaling, but I can’t always sit down and type everything out. In the past, I’ve done a lot of handwritten journaling, especially for dream journals, but I like my journals to be searchable so I can go back and find topics or ideas I’ve written about. When I handwrite, I end up having to transcribe it later, which is time-consuming. So I tried using voice notes, but those also need to be transcribed.

Even after transcribing, voice notes often don’t translate well because when recording audio, I’m not thinking about punctuation or formal writing—it’s more conversational. The transcription ends up full of errors, especially if I’m mumbling in the middle of the night while recording a dream journal. Add to that the noise of kids or dogs in the background, and the result is a jumbled mess with transcriptions that don’t come out right.

The Refine My Words GPT (https://chatgpt.com/g/g-xg4rSn6nf-refine-my-words) was born out of a necessity then.

I noticed that when I used the mic button with ChatGPT to record audio, the transcriptions were much better but still not perfect. That’s when I came up with the Refine My Words GPT. I instructed it to take whatever I write, refine it while keeping my original tone and intent, and rewrite it. This additional layer smoothed out the weirdness of direct audio-to-text transcriptions, fixing errors where the transcription wasn’t quite right. It typically corrected things very well.

I also found myself wanting timestamps for every entry, but I had to remember to speak the timestamp before recording what I wanted to say. So I added an action to this GPT to automatically get the current date and time in Central Standard Time (or any other time zone if specified). Now, every time I record a journal entry, it automatically appends the timestamp, refines the transcription, and outputs a nearly copy-paste-ready entry for me to put into my Obsidian journal.

This process allows me to make journal entries on the go—while driving, in noisy environments, or in the middle of the night when I’m half-asleep. I’m able to capture a lot more thoughts than before, without needing to sit down and write or type. It removed all the extra steps of recording audio, transcribing it, and then cleaning it up. Now, my journal entries are much more robust.

The drawback to this that I have found is that GPT can get a little more aggressive then I would prefer in it’s refinement of the transcribed audio. Perhaps after some more prompt refinement this will be able to be corrected. However, even when it does this, it usually does not greatly impact the flow or idea of what was recorded.

I made this GPT public and included instructions on how to use it. Feel free to use it if you find it helpful, as I certainly have. If you have any suggestions for improvements or encounter any issues, let me know. This can be a living, evolving GPT to enhance the journaling experience.