Hook Splitter

Community

mrpaidsocial

hook-splitter

Split ONE long recording that contains many separate hooks/takes into a set of standalone videos — one per hook — and tighten each so there is no dead air, no stumble, and nothing said twice. Use whenever the user hands over a single long file and wants the pieces cut out: "split these hooks out", "there are 20+ hooks in here, make them each their own video", "cut these into individual clips", "one video per example", "chop this composite up", "tighten these and send them to me". Also use for the follow-up rounds — reading the reviewer's timeline notes back off the review canvas and applying them. Built for the prompt→result demo shape (the creator sets up a prompt, plays an AI-generated video, reacts) but the machinery is general: any long recording of repeated takes. NOT for styling a single finished reel with banners/captions (reel-recut), NOT for a brand-deal graphics edit (branded-ad-edit), and NOT the canvas itself for a single cut (video-review-canvas — this skill ships its own 21-up gallery variant).

Overview

Publisher	mrpaidsocial
Repository	`mrpaidsocial-video-editor-agent`
Skill name	`hook-splitter`
Stars	114
Forks	34
Bundled files	17
Links	Skill source Repository

Markdown instructions
A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.
Works with any LLM
AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.
17 bundled files
Scripts, templates, and references the model can read while it works. Files are read-only and never executed.
Open source
Published by mrpaidsocial on GitHub. Read the source before you install it.

Installation

Install the Hook Splitter AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

Open the app and go to Plugins → Skills.
Choose "Install from GitHub".
Paste the skill folder URL below and confirm.
https://github.com/mrpaidsocial/mrpaidsocial-video-editor-agent/tree/main/.claude/skills/hook-splitter
Enable the skill in any chat where you want it available.

Store

Installed

Skills

Add skills

Import one or more skills from a public GitHub folder or a local .skill or .zip archive, or SKILL.md file.

From GitHub URLFrom .skill, .zip, or SKILL.md

GitHub repo or folder

https://github.com/mrpaidsocial/mrpaidsocial-video-editor-agent/tree/main/.claude/skills/hook-splitter

Points to a single skill folder, or any folder that contains multiple skills. Public repos only.

CancelContinue

Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills

git clone --depth 1 https://github.com/mrpaidsocial/mrpaidsocial-video-editor-agent.git /tmp/mrpaidsocial-video-editor-agent
mkdir -p .claude/skills
cp -r /tmp/mrpaidsocial-video-editor-agent/.claude/skills/hook-splitter .claude/skills/hook-splitter

Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Hook Splitter in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Hook Splitter on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Hook Splitter is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

Loaded only when it is needed
The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.
Switch models mid-chat
Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

hook-splitter — one long recording → N tight standalone hooks

Cuts a composite into one MP4 per hook, removes every gap and flub, and delivers the whole set on a here.now gallery canvas where each video has its own comment box.

Built and proven on Rough Cut - Composite.mov (16:26, 21 hooks): 946s of hook spans became 627s across 21 videos, then three feedback rounds off the canvas.

The two rules that matter most

1. Never trim copy for pacing. Remove dead air, stumbles, aborted takes and repeats. Do NOT shorten what the speaker actually said to hit a runtime. If a hook runs long because the setup is long, ship it long and say so — the trim is the user's call. (Approved copy on a paid deliverable is never trimmed for pacing.)

2. Transcribe the RENDER, not just the source. This is the single highest-value step in the pipeline. Whisper smooths a repeated take into one clean line on a full-file pass, so real doubles are invisible in every source-side transcript — but they show up immediately in a transcript of the cut. Four of the flubs the reviewer caught on the proving run were of exactly this kind. scripts/qa.py exists to catch them before the reviewer does.

Pipeline

Everything is driven by one hooks.json (see assets/hooks.example.json). Scripts take its path as $1 and default to ./hooks.json.

#	Script	What it does
1	`probe.py`	reads the source, extracts one wav per audio stream, computes 10 ms dB envelopes, and tells you which stream is the mic and which is the system capture
2	`video_regions.py`	the "a video is playing" mask (see below)
3	`detect_loops.py`	where each played video restarts (players loop)
4	`build_mix.py`	`work/mixed.mov`: video stream-copied, audio tracks summed, loop tails muted
5	`plan.py`	`cuts.json` — speech map, gap collapse, your hook table + kills
6	`render.py`	`final/hook-NN.mp4`, one per hook
7	`qa.py`	re-transcribes every render and scans for repeats (never skip)
8	`build_canvas.py`	the gallery canvas → publish with `here-now`
9	`map_notes.py`	round 2+: pull the reviewer's notes back and map them to source time

The spine: a clean "video is playing" mask

A screen recording made with a mic plus system-audio capture has two audio streams, and the system one is digital silence whenever nothing is playing. That gives an exact, free mask, and the entire edit hangs off it:

inside a video region → never cut. That is the only silence the brief allows.
outside one → the speaker is talking; every gap collapses to a fixed residual.

probe.py identifies the streams by their silence fraction (the system track is

50% digital silence; the mic never is). Verify alignment before trusting it — transcribe the same 45s window from both near the start and near the end. It has been exact so far, but a drifting mask would corrupt all 21 outputs silently.

Only one audio stream? Then there is no free mask. Mark the video regions by hand in hooks.json ("video_regions": [[13.2,16.45], …]), reading them off scripts/probe.py --map output plus the transcript. Everything downstream is unchanged. Do not try to infer them from the mic track alone — the ad audio bleeds in ~22 dB down and sits right on the speech threshold.

Delivery

build_canvas.py + the here-now publish script. It is a 21-up gallery, not the single-video canvas: the comments collection carries a hook field so each video gets its own box, and a note stamps that player's current frame.

Reply LEADS with the canvas URL, every round (see video-review-canvas).

Bump version every round so answered notes stop cluttering, and version-stamp the filename of every hook whose bytes changed (hook-04-v2.mp4) so no one gets a cached old cut. Delete the superseded files; verify the old name 404s.

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Hook Splitter AI skill do?

Why use Hook Splitter on TypingMind?

Because you install it once and use it with any model. Hook Splitter is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Hook Splitter in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/mrpaidsocial/mrpaidsocial-video-editor-agent/tree/main/.claude/skills/hook-splitter. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Hook Splitter?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Hook Splitter?

As many as you like. As long as a model supports skills, you can use Hook Splitter with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Hook Splitter AI skill free?

It is published on GitHub by mrpaidsocial. Check the repository for licensing terms. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Nemo Curator

NousResearch

OrganizationPopular

Curate LLM training data: dedupe, filter, PII redaction.

Obliteratus

OBLITERATUS: abliterate LLM refusals (diff-in-means).

Prompt Lookup

Activates when the user asks about AI prompts, needs prompt templates, wants to search for prompts, or mentions prompts.chat. Use for discovering, retrieving, and improving prompts.

Banner Design

Design banners for social media, ads, website heroes, creative assets, and print. Multiple art direction options with optional generated or supplied visuals. Actions: design, create, generate banner.…

Hook Splitter

Overview