YouTuber Cuts Video Production From 22 Hours to Under 60 Minutes With AI

A creator who has published weekly YouTube videos for 14 years says an AI editing workflow has cut his production time from 14–22 hours per video to under an hour — with long-form videos now going live in around 40 minutes. Justin Brown, Co-Founder of Primal Video, laid out the full process on Social Media Examiner's AI Explored podcast, and it's a working pipeline, not a demo.

The setup: the Claude app running Code, or the ChatGPT app running Codex, editing files locally on the creator's hard drive rather than in the cloud. Local processing uses fewer tokens (paid units of AI usage), which keeps costs down. Brown's current go-to is the latest ChatGPT model in Codex; he also gets good results from Google's Antigravity, though he considers it a fair bit behind.

What does this change for creators who can't edit?

The barrier to entry, Brown said. You no longer need to learn Adobe Premiere, Final Cut Pro, or DaVinci Resolve, or understand audio levels and music composition. You learn prompting instead — writing instructions for the AI — and focus on storytelling and the message. He rejects the claim that AI is bad at editing, arguing failures come from typing "edit my video" and hoping for the best. An AI editor, like a human one, needs clear guidelines.

His workflow runs in ten steps:

• Train the AI on standout videos. Define the video's goal — awareness, leads, followers, or sales — then feed the AI links to leaders in the niche. It reverse-engineers pacing, title frequency, and music use before touching your footage.
• Shoot for the edit. Record one continuous file, always use the last take, pause between statements, and restart full sentences when a line goes wrong. The AI follows spoken cues like "actually, use the first take" more reliably than human editors did.
• One project folder per video. Subfolders for raw footage, B-roll (overlay shots laid on the main camera track), music, photos, and graphics. Prompt the AI to inventory everything first.
• Get an accurate, time-stamped transcript. Brown calls this the most important step. Timing matters more than words, so cuts land between words instead of clipping them. He warns the AI can "gaslight" users, claiming it transcribed everything while missing sections. His pick for long-form: Deepgram, a cloud service where a $20 prepaid balance has lasted months.
• Two passes for long-form. A story pass builds the A-roll (the main talking-head track) first; a polish pass adds titles, graphics, and music. Ask for a timed script of what was cut and kept to review cheaply.
• HyperFrames for motion graphics. The open-source framework from HeyGen places text behind the speaker, generates animated graphics from the transcript, and tracks the mouse in screen-recorded tutorials. Brown doesn't use HeyGen's AI avatars — it's strictly for enhancing real footage.
• Music via MCP connection (a protocol linking external tools to the AI). Epidemic Sound and Artlist plug in directly; the AI selects and places tracks, and swaps them on requests like "something a bit more upbeat."
• ChatCut for review. The free Mac and Windows tool shows a visual timeline without manual editing, previews without rendering, and lets Brown make small tweaks the AI learns from.
• An orchestrator chat manages the whole edit. Named as "the manager," it spins up sub-agents for captions and graphics, checks their work against standing rules, and reports back only when the video is ready.

One concrete budget note: without a B-roll percentage specified, Brown tells the AI he wants videos "engaging and clear," or sets a mix such as 45% B-roll and 55% on-camera subject. Bad cuts trace back to one of two causes, he said — an unclear instruction or footage that made the decision hard — and each fix becomes a permanent rule.

Practitioners wanting to test this can run a simple edit on existing footage and ask the AI to walk them through setting up Deepgram step by step. The open question is whether these time savings hold at scale; Brown's 40-minute long-form turnaround suggests they already do for his team.