Table of Contents

This article describes how I reduced the time spent writing travel blog posts from 3–4 hours to about one hour by rethinking my workflow and using AI-powered tools. By switching from typing to voice notes, automating transcription, and leveraging ChatGPT to clean up text and generate Hugo-ready Markdown, I streamlined content creation significantly. The result is a faster, more enjoyable blogging process and several extra hours of free time each week.

Introduction

How to save 3 hours per week of life with AI?!
In this article, I’m going to share how I managed to decrease the amount of time I spend preparing a trip review from a couple of hours to about an hour (3-4 times). It’s not an article on text generation, fine-tuning, or other AI techniques, but more on how to chain various tools to make processes more efficient.​ In my blog, I’m trying to document my travel experience, but due to a lack of time, I managed to cover only one of six trips this year. Btw, don’t forget to check out our fantastic experience in Bhutan - start from day-1!!​

Original Content Creation Workflow

Usually, to prepare an article, I go through the following steps:

Text Preparation Stage

Firstly, I just dumped my thoughts on a trip in a Google Doc. At this step, I don’t spend much time thinking about grammar, structure, or links.

Secondly, I go through the text and come up with a proper structure, review how cohesive it is, and use Grammarly to check my grammar.

Lastly, I added links & references that I mentioned during the narration.

This stage usually took about 2-3 hours.

Image Preparation Stage

Firstly, I export all trip-related images from my phone to my Mac, pick the most notable / interesting, and, lastly, sort them according to the day of the trip.

After that, I convert the images from HEIC to JPEG, and create a thumbnail image

Lastly, I upload these images to a dedicated S3 bucket.

That’s a pretty straightforward step, and I spent about 30-40 mins.

Markdown & Hugo Formatting

At this stage, I have to convert the text to a Hugo-digestible Markdown page. It’s not a difficult process, but quite a booking stage, where I have to pick relevant keywords, tags, etc.

This stage takes about 15-30 minutes.

Deployment

Lastly, I deploy the website to AWS and kick off invalidation of a CloudFront distribution. It’s still a manual step, but I’m going to fix it soon.

Another 5-10 mins!

The Problem Summary

To conclude, it’s not an easy process that can easily take 3-4 hours of my time per post. Recently, I managed to significantly improve the process and reduce the per-post time to roughly an hour.

Voice-First Content Creation

Firstly, I realised that I can express my thoughts much faster by my voice than by typing. Secondly, it’s not hard to see that each article takes only 5-6 minutes to read, so I thought it would be far easier to talk through the experience and then convert my voice to text. Thus, I decided to test this hypothesis during our last trip to Taiwan. ​ After each day, I record a voice note on my phone about the day. Each note length is roughly only 15 minutes! Once we returned, I pretty quickly sorted out how to leverage Google Cloud to translate voice to text: create a bucket with files, upload voice files (I converted m4a to flac via ffmpeg), and kick off the transcription! And, voila, post ready! Obviously, I have a lot of garbage words, so I leverage ChatGPT to clean up my lexicon and highlight grammar issues that I can fix during the review.

NOTE: While I was writing an article, I discovered that Google API isn’t needed, and the iPhone can translate your voice memos to text easily.

Automating Hugo with ChatGPT

​But that’s not it! I was quite annoyed with writing a markdown for a post every time, but ChatGPT can easily help with it. I spent some time tweaking the prompt and managed to get perfect results. To avoid saving the prompt, I created a custom GPT to format a pure piece of text into a Hugo post. Now I have 30 minutes of my life back every week!

Conclusion & Next Steps

To conclude, with the above automations, I managed to bring down time per post from 3-4 hours to about an hour (or even shorter). The next step is to vibe-code an application to join the above steps together. I will provide text and images, and the app will combine them into a hugo-disestable post, which will be published after a review!

Stay tuned!