Skip to content

patchmyday/brain-dump-workflow

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Brain Dump Workflow — PatchMyDay Video #1

"Your best ideas happen in the shower and die before you sit down."

First video for PatchMyDay channel. iPhone Shortcut that captures voice to timestamped Markdown in iCloud. Built entirely with AI assistance (Hermes Agent + ElevenLabs + FFmpeg + Puppeteer).

What's In This Repo

This documents the ACTUAL production process — every iteration, every dead end, every tool used. Not a template. A real build log.

Path What It Is
production-log/ Blow-by-blow of how this video was made, including failed approaches
scripts/ EN script (v1 + v2), CN script, section JSON configs
subtitles/ Proofread SRT files (EN + CN)
scenes/ HTML5 Canvas scene source code (v1 through v4)
audio/ ElevenLabs voice config, section splits, concat files
xiaohongshu/ Chinese cross-post content (script + 小红书 graphic text post)
PRODUCTION-PLAYBOOK.md Reusable playbook extracted from this production

Video Iterations

Version What Changed Result
v1 Manim animation (Python) Too slow, limited visual options
v2-lottie HTML5 Canvas + LottieFiles characters Fun but generic, no real screenshots
v3-screenshots Real iPhone screenshots composited into scenes Good but not synced to voice
v4-synced Each scene duration-matched to voice section Correct sync, final structure
v5-proofread SRT corrections (15 fixes) burned onto clean v4 base FINAL

Final Output

  • Duration: 4:34 (274s)
  • Master: 1920x1080, H.264, CRF 23, ~21MB
  • Compressed: 1280x720, CRF 26, ~6.3MB (Telegram-friendly)
  • Voice: ElevenLabs Adam voice, Multilingual v2
  • Subtitles: Burned-in, Arial 22pt white + black outline
  • Music: Background lo-fi mixed at -18dB (volume 0.12)

Tools Used

  • Hermes Agent — AI assistant that wrote scripts, generated scenes, ran FFmpeg, proofread
  • ElevenLabs — TTS voice generation (Creator tier, Adam voice)
  • Puppeteer — Rendered HTML5 Canvas scenes to frame sequences
  • FFmpeg — Everything: concat, mux, subtitle burn, compression, audio mixing
  • Whisper — Generated initial SRT from voiceover audio
  • Manim — First attempt at animation (abandoned for HTML5 Canvas)
  • Node.js + Chromium — Puppeteer runtime for headless rendering

Brand

PatchMyDay — Cybersecurity + AI Automation. Bilingual EN/CN. Fireship-style (no face cam, fast cuts, screen recordings + motion graphics).

About

Voice Brain Dump workflow using iPhone Shortcuts. Capture ideas by voice, auto-save to iCloud as .md for AI agent processing. Part 1 of a series.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors