How I Built My Own AI Tool for My YouTube Videos with Claude Code

A while ago I watched a video about Claude agents, and it made me curious enough to try Claude for a month, even though I was already using ChatGPT every day. When I noticed Claude Code, the question became: what could I build that would actually make my life easier? The answer was my videos. After recording, I use Whisper to create a transcript, and then I spend a lot of time fixing misheard words and hallucinations, working out titles and chapters, and looking for good moments for Shorts. So I started building the AQL Video Processor. Not a program that makes my videos for me, but one that helps with the repetitive work so I can keep making my own decisions. In this video I talk about how the project grew, milestone by milestone: • Planning the workflow with ChatGPT before asking Claude Code to write any code • A Python command line tool that reads Whisper JSON and flags low-confidence words, suspicious gaps, repeated words and names from my vocabulary list • Preparing review packages for an AI to check, and processing its corrections back into subtitle files • Validation and automated tests, so the tool doesn't create more work than it saves • Generating YouTube metadata and finding possible moments for Shorts I have programming experience, but I'm not a professional Python developer, so this was very much a learning project. My biggest lessons: know what you want before you ask an AI to build it, work in small milestones, and don't blindly trust the output, whether it's reviewing a transcript or writing code. Every finished milestone felt a bit like completing another quest. In my next video about the AQL Video Processor, I want to show you how I actually use it, from a Whisper transcript to reviewing subtitles, generating metadata and finding moments for Shorts. Are you using Claude Code? Are you building your own tools with it, or using it for something completely different? 00:00 Trying Claude and Claude Code 01:09 Finding Something to Build 01:42 Transcripts, Metadata and Shorts 03:47 Planning the Project with ChatGPT 05:39 GitHub Repository and Version Control 06:07 Milestone 1: Reviewing Whisper Transcripts 09:10 Milestone 2: AI Review Packages 11:10 Corrections, Validation and Tests 13:47 ChatGPT and Claude Code Together 15:57 What I Learned 17:48 Was It Worth It? 18:57 Recap and Next Video #AriQuestLog #ClaudeCode #Claude #ChatGPT #Whisper #Python #AIAssistedCoding #AQLVideoProcessor #YouTubeWorkflow #ContentCreation