Beatscope Visualizer logo

Beatscope Visualizer

Community
chosuicide
beatscope-visualizer

Use BeatScope music facts, grouped edits and seek-safe motion for video with optional references. Read AGENT.md first; no fixed visual style.

Overview

Publisherchosuicide
Repositorybeatscope
Skill namebeatscope-visualizer
Stars
79
Forks
5
Bundled files
36
LicenseMIT
Links
  • 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.

  • 36 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by chosuicide on GitHub. Read the source before you install it.

Installation

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

1

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.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/chosuicide/beatscope/tree/main/beatscope/agent_skill
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

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/chosuicide/beatscope.git /tmp/beatscope
mkdir -p .claude/skills
cp -r /tmp/beatscope/beatscope/agent_skill .claude/skills/beatscope-visualizer
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Beatscope Visualizer 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 Beatscope Visualizer 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 Beatscope Visualizer 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.

Query music and execute an edit

Read AGENT.md. Resume a matching project within scope, or create a new work. Query summaries, not whole maps: node query.mjs 81 103 --accents --stages. Exact meanings: BEATSCOPE.md/references/schema.md. Visual choices: references/directing.md. Identity, authorization and cached evidence: references/handoff-rules.md. End-to-end sample: references/complete-example.md. JSON imports need Node22+ or current Chromium over HTTP with JSON/JS MIME. Rendering dependencies and limits live in references/rendering.md. The brief's stages are resolved editor stages; automaticSections and measuredAnchorCandidates remain factual suggestions. cueEdits.shown is a bounded inspection list, not the full cue schedule. Resolve the plan for execution.

Facts

js
import * as music from './visual-state.js';
import {createTrack} from './beatscope-runtime.js';
const track = createTrack(music.RHYTHM_MAP);
const candidates = (start,end,budget) => typeof music.getResponseEvents === 'function'
  ? music.getResponseEvents(start,end,budget) : track.responseBetween(start,end,budget);
function frame(frameIndex,fps,excerptStart) {
  return music.getVisualState(excerptStart+frameIndex/fps);
}

State includes beat/bar phase, bands, onset/accent and optional structure/meanRms. The budget returns possible measured anchors, not musical functions or cut rules. Use ordinary beats and continuous activity alongside accent anchors.

Edit score: authored by Agent, executed by code

Resolve the supplied editor plan once before authoring:

js
import plan from './edit-plan.json' with {type:'json'};
import {resolveEditPlan} from './edit-plan.js';
const {stages,cues} = resolveEditPlan(music.RHYTHM_MAP, plan);

These stages define contiguous passage boundaries; add focus/change/rhythm to them in your score. Unchanged cues are measured candidates; manual:true cues are added/moved response anchors. Deleted cues are omitted. Use resolved cues for authored responses so overrides take effect. Keep sourceTime for citation; do not also fire the old time of a moved/deleted cue. A cue can drive several layers, a cut or a returning sequence. It does not specify one isolated action.

Keep one job-local score.json outside this package. Reuse it when present; preserve accepted shots, effects, asset bindings and unrelated timing during scoped edits. If it belongs to another song, reuse suitable visual definitions separately and author timing against this package. Users do not fill it in. For an MV set intent:"music-video". First author the musical passages and their visual rhythm, then shot setups/sequences, then motion inside shots. Import createEditScore from edit-score.js. The score contains:

  • stages: adjacent whole-song {id,start,end,focus,change,entry}. entry names a shot at that boundary; carry:true explicitly continues an earlier shot. Each MV stage has rhythm blocks covering it exactly, with {start,end,mode,music,reason,anchors,action?}. Anchors use the same time API. mode hold permits one shot identity and requires action describing its motion and destination; switch requires at least one identity change; alternate requires at least two. music describes the actual musical passage; reason connects its visual choices. anchors are increasing music cues inside it. These checks enforce declared edits, not taste or a universal cut frequency.
  • shots: reusable {id,relation,setup,values,tracks,clock,duration}. setup is renderer configuration; values are static named properties; optional tracks animate those properties in LOCAL seconds. relation describes why shots belong together. The renderer still provides actual footage/geometry/materials.
  • cuts: optional individual {at,shot} entries.
  • groups: {id,stage,anchors,sequence} binds one authored combination to several real music events. Sequence cells name shots; null holds the current shot while other music-linked tracks may act. Lengths match unless repeat:true explicitly repeats the supplied combination. No built-in combination or cut frequency.
  • moments and global tracks: shared music-time anchors and motion, as below. Global property writers cannot overlap shot-local property writers.

A shot clock uses {mode,offset:0,rate:1}: continuous (MV default) advances since first visit including offscreen time; visible (other scores' default) advances only while onscreen; restart resets on every explicit visit. Choose visible/restart when a pause/replay has a purpose. Returning can continue an action, pause it or deliberately replay it. duration bounds sampled local tracks; view.localTime reports raw clock and view.sampleTime reports bounded time. Footage decoding/looping remains the renderer's job.

Abstract example; replace relationships, views and anchors for the actual job:

js
import * as music from './package/visual-state.js';
import {createEditScore} from './package/edit-score.js';
const score = {
  intent:'music-video',
  stages:[{id:'first',start:0,end:music.RHYTHM_MAP.duration,entry:'wide',
    focus:'chosen subject',change:'detail changes our reading of the whole',
    rhythm:[{start:0,end:music.RHYTHM_MAP.duration,mode:'alternate',
      music:'example repeated attacks; replace with actual passages',
      reason:'whole and detail reveal the same subject',anchors:[{beat:1},{beat:3}]}]}],
  shots:[
    {id:'wide',relation:'whole subject',clock:{mode:'visible'},
      values:{'camera.z':10},setup:{sourceId:'chosen-source'}},
    {id:'detail',relation:'detail of that subject',clock:{mode:'continuous'},
      values:{'camera.z':2},setup:{sourceId:'chosen-source'}}
  ],
  groups:[{id:'phrase',stage:'first',
    anchors:[{beat:1},{beat:2},{beat:3}],
    sequence:['detail',null,'wide']}],
  tracks:[]
};
const edit = createEditScore(score,music);
function seek(localTime,excerptStart=0) {
  const state = edit.at(excerptStart+localTime);
  // Select state.view.id/setup; apply state.values through one adapter.
  // Use view.sampleTime for local motion/media; decode before final rendering.
  return state;
}

A view is a complete shot setup, not automatically a photo card or a slide. Repeated visits do not require a new object. Combine cuts, holds and continuous shots according to the reference and musical phrase. at(t) is independent of query order. cuts exposes resolved times; frameWarnings(fps) identifies visits with no output frame. These warnings never silently shift measured anchors.

Optional picture tools

Reuse existing framing/bindings when suitable. Read references/picture-tools.md only if needed: frame2D fits a 2D target, frame3D fits a perspective target, createPictureBindings connects compiled shots/properties to native setters. They add no production step. Read the relevant API section for the chosen renderer; helpers provide no style or automatic camera path.

Shared motion

createChoreography(score,music) remains available for motion-only work. moments are named anchors; tracks are unique {id,keys} property writers. Keys: {at,value,ease}; value is a number, numeric vector, string or boolean. Outgoing ease: linear, smooth, in-cubic, out-cubic, hold. Discrete changes need hold. The destination value is exact on its key; last values persist. Anchor: seconds, {beat:index} (zero-based array index), {onset:id} (measured instant), or {moment:name}. Object anchors accept offsetBeats then offset seconds. Real local beat spacing preserves fractional onset phase; no quantizing. Several tracks can share one moment with different preparation/aftermath. Optional pace:{from,to,every,step,mode} adds a scalar/vector increment per beat interval; step advances discretely, travel continuously. Final advance persists. Optional activity:{band,amount} adds selected low/mid/high/all activity to that same property writer. No default pump, shake, particles or slow easing.

For travel that must survive cuts, add one GLOBAL progress track to score.tracks:

js
{id:'world.travel',keys:[{at:0,value:0,ease:'hold'}],
 pace:{from:{moment:'enter'},to:{moment:'arrival'},every:1,step:1,mode:'travel'}}

Define enter/arrival from the actual job's anchors. This value interpolates real beat spacing, keeps advancing across cuts and retains its result at arrival. The adapter uses the SAME value for camera path and supporting depth layers; shot setup chooses the viewpoint. Do not substitute view.sinceCut for world progress: it restarts each cut. Use sinceCut only for deliberate cut-local settling. Short impact/arrival easing may finish while travel continues. Once the journey ends, author the next process or an intentional rest; continuous agitation is not required. If the candidate grid is unsuitable, use a linear global track between chosen measured seconds instead of claiming beat-locked motion.

One technical check

Run node package/music-brief.mjs --score score.json --mv once after MV changes. It checks spans, anchors, cut conflicts, writers, return clocks and reverse query order and declared rhythm without rendering. It reports switches and longest visit per passage, plus unsampled visits at30fps; inspect with the actual output fps. It does not prove beauty, reference fidelity or listening. For a dense interval node music-brief.mjs --groups start end maxGap [minEvents] reports numerical onset runs (bounded output, explicitly marked truncation). They are not confirmed phrases or an automatic cut schedule. Programmatic groupOnsets(map,start,end,{maxGap,minEvents}) returns complete runs when needed.

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 Beatscope Visualizer AI skill do?

Use BeatScope music facts, grouped edits and seek-safe motion for video with optional references. Read AGENT.md first; no fixed visual style.

Why use Beatscope Visualizer on TypingMind?

Because you install it once and use it with any model. Beatscope Visualizer 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 Beatscope Visualizer in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/chosuicide/beatscope/tree/main/beatscope/agent_skill. 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 Beatscope Visualizer?

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 Beatscope Visualizer?

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

Is the Beatscope Visualizer AI skill free?

Yes. It is published on GitHub by chosuicide under the MIT license. 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

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇