Higgsfield's Katana rebuilds a viral edit shot for shot using your own footage, and a custom Claude skill turns Claude into the creative director that plans it for you.
Higgsfield's Katana tool rebuilds any viral video edit shot for shot using your own footage inside Claude, and a custom Claude skill turns the hardest part, picking the edit and writing the story, into a repeatable creative-director workflow.
Higgsfield's Katana tool, accessed inside Claude through the Higgsfield MCP, rebuilds any existing video edit shot for shot using your own face, product, or footage. It matches the original's cuts, timing, colors, and titles exactly, and keeps the real editing work (text, flashes, camera moves) as code rather than AI video, only generating the specific shots your footage is missing, so you only spend credits on what's actually new. Because most people connect Katana and then stall on which edit to pick and what to say, the creator built a Claude skill, Katana Director, that interviews you about your business and audience, chooses the edit, writes the story, and hands the finished plan to Katana automatically.
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 →
Jad opens by showing his original motion graphics lyric video (160K+ views) next to a Higgsfield Katana remake with the exact same cuts and beat timing, but with himself now in the shot.
He explains his no-sponsor-until-now stance, why Higgsfield (a tool he already used for his avatar, sneakers, and clothing brand carousels) earned the spot, and his promise to keep giving the underlying skill away for free.
Katana takes an edit that already works and rebuilds it shot for shot with the user's own face, product, or clips, matching the cuts, motion, color, titles, and audio of the reference.
Katana runs through the Higgsfield MCP connector, powered by the same engine behind Claude Motion. Users either pick a community preset plus up to 5 photos, or upload their own edit plus up to 40 reference images.
The flashes, titles, color grading, and camera moves are generated as code, with Claude writing it frame by frame, not as AI video. AI generation is reserved only for the shots the user's footage can't cover.
Segments built from code or the user's own footage cost zero Higgsfield credits. Only the newly generated AI shots consume credits, so the cost line shows exactly what's being paid for.
Most people connect Katana and then stall on which edit to pick, what the video should say, and what the hook is. The Katana Director skill turns Claude into a creative director that handles that planning automatically.
Jad demonstrates the workflow on a real example: an ad for Ember & Bun, a burger spot from an earlier carousel video, rebuilt as a new Katana edit.
He scrolls the Katana presets library of edits other creators have published, then swaps his own face into one of the popular ones to show the frame-by-frame recreation in action.
Katana supports three paths: generate fully static AI images (cheapest), mix static images with short AI video clips, or skip generation and bring an existing video for Claude to edit directly in the motion-graphics style.
Jad hands Claude a video he already has and asks it to edit the footage together in the motion-graphics style, with no new AI-generated shots required.
Step by step: download the skill zip from the Skool community post, then in Claude go to Settings, Customize, Skills, Upload a skill to install it.
After pasting the skill's starter prompt, Claude interviews him like a creative director: the business, the audience, where it will run, organic or paid, the offer, and the brand name, then generates character and brand style sheets.
Using only a selfie and an unrelated motorcycle photo, with no written prompt, Jad walks through Katana planning the shots, generating them, and assembling the final styled edit in real time.
First-time video businesses, a food shop, clinic, or gym, need only a few photos or a phone clip. Brands can test three hooks on one edit, and marketers can run a V1-to-V4 batch and keep the winner.
Jad closes by pointing viewers to the Higgsfield link and his Skool community for the free Katana Director skill, and commits to a new Claude skill every week.
Katana rebuilds any edit shot for shot from your own footage, keeps real production work in code instead of AI generation, and a saved Claude skill turns the planning into a repeatable creative-director workflow.
“This is motion graphics on the left, that's the lyric music video from my motion graphics video, the one that did over 160,000 views, and on the right, Katana rebuilt it shot for shot.”
“The flashes, the text, the colors, the camera moves, that's code. Claude writing code frame by frame, that's motion graphics.”
“A skill is a set of instructions you add to Claude that teaches it how to do a specific job. Once it's installed, Claude follows the playbook every single time.”
“What should the video even say? What's the hook? Most people have no idea.”
“Try to find an editor that you love and open Claude. It'll make yours.”
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.
Jad opens with a split screen: his own 160K-view motion graphics video on one side, an AI remake of it with himself in the shot on the other, every cut and beat preserved. The rest of the video is the walkthrough of how that remake got made, and the Claude skill he built so the planning doesn't fall on you.
The three input paths Katana supports, ordered from cheapest and most AI-generated to most footage-driven and cheapest in credits.
The question sequence the Katana Director skill runs before it picks an edit and writes the story, so the output is scoped to a real brief instead of a generic template.
“Find an edit you love, open Katana in Claude, and make it yours.”
A short, direct close that repeats the link-in-description ask made at the top of the video and points back to the free community skill, without a hard sales push.
00:00
00:19
00:2700:38
00:49
01:00
01:11
01:22
01:33
01:44
01:56
02:0702:18
02:2902:37
02:5103:02
03:13
03:20
03:38
03:4603:5704:08
04:19
04:30
04:4104:52
05:02
05:1005:26
05:37
05:50
06:03
06:1006:21
06:32
06:40
06:53
07:00
07:1207:27
07:38
07:49
08:0008:14
08:21
08:28
08:4009:00
09:02
09:18
09:29
09:40
09:53
10:02
10:1310:22
10:35
10:46
10:59
11:03
11:18
11:30
11:41
11:52
12:03
12:14
12:25
12:32
12:47
12:59
13:1313:18
13:32
13:4313:54
14:05
14:16
14:27
14:38Add Modern Creator as a preferred source and Google shows you more of our breakdowns in Search, Top Stories, and AI Overviews. It only changes what you see, and you can undo it in your Google settings anytime.
16:12A creator walks through the Claude Code skill that transcribes, cuts, researches, builds motion graphics, grades color, and finishes his videos end to end, and explains what it actually costs to build and run.
October 10th
14:57A SaaS app, a restaurant, and a clothing brand each get a six-slide motion-graphics carousel, coded and rendered locally by Claude instead of any AI video model.
October 7th
24:38A creator rebuilds a viral AI music video twice. Once with paid tools, once on nothing but a Claude subscription. Then he hands out the exact skill for free.
October 9th
15:46A creator maps six tiers of Instagram carousel quality, from copy-paste Canva templates to animated avatar carousels, and gives away the Claude skill that builds all of them for free.
October 6th
15:47A breakdown of turning 10,000-character prompts, local audio synthesis, and automated critique loops into high-end motion graphics with Claude and zero paid APIs.
October 4th
15:46A RoboNuggets creator walks through the three-step Claude Sonnet 5.5 workflow he uses to cut, animate, and direct his own videos end to end, with no separate AI video-generation model involved.
October 2nd