How I Fully Automated My Video Editing (Claude Code)
A 12-minute screen-recorded walkthrough of a three-tool pipeline that takes raw footage to finished export without touching a timeline.
April 11thA 22-minute live tutorial where Jason Cooperson builds and demos a Claude Code pipeline that cuts, graphics, captions, and exports a video, using the intro you just watched as the live proof.
Claude Code can fully automate video editing if you wrap it in the right scaffolding: a WhisperX rough cutter, a HyperFrames graphics engine, and a partial-render system that makes the second pass loop fast enough to actually use.
Jason spent two weeks building a Claude Code project folder that turns raw footage into a fully edited video. The pipeline is seven stages: intake, rough cut via WhisperX word-level timestamps, graphics via HyperFrames, a second-pass refinement loop, auto-captions, background music, and export. The key engineering insight is a partial-render system that re-renders only the segment you changed, making the back-and-forth prompting loop fast enough to actually use. The demo edits the video own intro live, going from a 4-minute raw clip to a 47-second polished cut with motion graphics on the first try.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →
Opens with the claim that this video itself was edited by Claude, then demos motion graphics and captions from the finished cut.

Explains the project is a folder of Claude Code files. Walks through the 7-stage pipeline and the two paths: build-it-yourself vs. download the zip.

Opens Claude Code, creates or downloads the video-editor project folder, installs HyperFrames, WhisperX, and FFmpeg dependencies.

Drops the 4m10s raw intro clip into Claude, runs the rough-cut skill, waits 10-15 minutes, reviews the 47-second output. No manual adjustments needed on first try.

Sends the run-graphics prompt, waits 22 minutes, previews the AI-generated graphics pass.

Iterates on individual graphics via natural language prompts. Partial re-render takes about 3 minutes per change. Full second pass takes 2 hours.

Adds animated word-pop captions for short-form; reuses existing WhisperX transcription, applies Coolvetica preset template.

Copies music file path, pastes into Claude, specifies -23dB after playback, Claude mixes it in.

Exports final video to downloads folder while keeping the project folder intact for future re-edits.

Reflects on time savings, claims this is the best Claude video editing system publicly available, pitches the Skool community.
The rough cut is fast and surprisingly good, but the gap between AI slop and a polished video is closed entirely in the second pass, where human direction drives iterative graphic refinements one segment at a time.
“I built the most cracked, dialed-in Claude video editing system on all of YouTube”
“This is the difference between AI slop versus like an actually good edited video”
“I spent basically two full weeks working non-stop. It nearly killed me.”
“As models continue to improve, if we ever get Fable back even, then this system just automatically makes it that much better”
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
The video you just pressed play on was edited entirely by an AI, and that is the first thing the creator tells you. Before any setup slides or pipeline diagrams, the proof is already running.
A linear pipeline where each stage has a dedicated Claude Code skill. The rough cut and graphics passes are the heaviest; the rest are quick.
Baked-in presets that determine the visual layout and graphics style for different distribution formats.
“If you want the full plug-and-play files with all the custom skills and all the workflows that I built, that will be the first link down in the description below.”
Repeated twice: once at the free-vs-paid fork early in the video and again at the end. Shows the Skool classroom page, zooms into the download button. Soft pitch; does not hard-close.
00:00
00:25
00:46
00:59
01:15
01:32
01:49
02:06
02:23
02:40
02:57
03:14
03:31
03:47
04:04
04:21
04:38
04:55
05:12
05:29
05:46
06:03
06:19
06:36
06:53
07:10
07:27
07:44
08:01
08:18
08:35
08:54
09:08
09:25
09:41
09:59
10:16
10:33
10:50
11:07
11:23
11:40
11:57
12:14
12:31
12:48
13:05
13:22
13:39
13:55
14:12
14:29
14:46
15:03
15:20
15:37
15:54
16:11
16:27
16:44
17:01
17:18
17:35
17:52
18:09
18:26
18:43
18:59
19:16
19:33
19:50
20:07
20:24
20:41
20:58
21:15
21:31
21:48
22:05
22:22A 12-minute screen-recorded walkthrough of a three-tool pipeline that takes raw footage to finished export without touching a timeline.
April 11thA 23-minute end-to-end guide to making AI-edited video from natural language -- no timeline, no drag-and-drop, just HTML rendered locally.
June 23rdAn 11-minute field report on three AI tools that survived a week of real production work.
June 12thA creator with 20 years of editing experience built a custom Claude skill that edits a 40-minute raw talking-head video in 12 minutes -- then gave it away free.
June 15thA 21-minute walkthrough of a six-phase Content Operating System that turns three days of client footage into a full month of distributed social content in one focused day.
June 13thHow a non-developer built a self-improving content factory that turns one source video into four platform-native formats and gets smarter every night.
June 11th