VividKit Crystal clear AI coding
VI
Choose Your Setup Get Started with AgentKit Continue with ClaudeKit Migrate from ClaudeKit to AgentKit Clean cutover from ClaudeKit What Is AgentKit? How AgentKit Works AgentKit Kits, Skills, Agents & Hooks AgentKit Runtime Support AgentKit Desktop App AgentKit Configuration AgentKit CLI Quick Reference AgentKit Skills Cheatsheet AgentKit Workflows AgentKit Statusline Update AgentKit CLI & Kits AgentKit Helper Clean reinstall of AgentKit AgentKit Troubleshooting What is ClaudeKit? ClaudeKit CLI Guide Command Finder Commands Cheat Sheet Workflow Recipes CLI Quick Reference Claude Mechanics Permission Modes ClaudeKit Hooks Coexistence ClaudeKit x Codex CK → Other coding tools How ClaudeKit Works Inside ClaudeKit Finding Your Unknowns Session Recovery Fix from Logs UI/UX Guide Accounts Switcher IDE Config Remote Control Codex App System Diagramming UI Review Gate Promotions Support VividKit

Quick Ref / Command

Marketing Kit

/ak:motion-video

Beat-synced motion video

Produce a 1920×1080, 30 fps HyperFrames motion-graphic video with AI voice-over, Vietnamese karaoke captions, SFX, and music cut to the beat, in one of about 60 named style profiles or a per-scene mix, with production-grade motion and fast 1:1 / 9:16 editions. It produces the files and never publishes them.

01

Pin the outcome

02

Scaffold the project

03

Generate and align audio

04

Fit beat and arrange music

Rule 01

Scene cuts land on music beats and reveals land on spoken words; the mix is measured, not guessed.

Rule 02

Every look is a style profile; brand names are aliases for a motion language and never bring logos, fonts, or assets.

Rule 03

On-screen facts must trace to the primary sources collected at the start.

Rule 04

Done means measured numbers pass, not that the render looks right.

Execution Map

Core lanes from input to output, aligned with the command's real execution path.

01

Start

Input, scope, route

  1. 1 Pin the outcome Record topic and sources, base style and any layers, duration, voice language (English voice-over with Vietnamese captions by default), ending, and signature. Ask only when the style or facts source is unknown. By default a director subagent writes the creative direction before any audio is generated; --poor runs everything in one session.
  2. 2 Scaffold the project Create assets/videos/<slug>/ and a plan, then copy the template scripts, example data, hyperframes.json, and the composition skeleton. A finished edition is never overwritten; a restyle is a new directory.
02

Work

Agent / skill execution

  1. 3 Generate and align audio Write facts and script from primary sources, then use multix for Gemini TTS voice, ElevenLabs SFX and music, and ElevenLabs forced alignment of spoken words.
  2. 4 Fit beat and arrange music fit-beat-grid.py reports BPM and a per-bar table; bars are spliced so drops land where the script needs them, and verify-arrangement.py checks every splice.
03

Verify

Gate, review, validation

  1. 5 Build timeline and composition build-timeline renders the ducked, loudness-normalized mix; index.html follows the composition contract and resolved style, then passes hyperframes lint, check, scene snapshot review, and the director's pass on every scene (focal point, depth, overlapping choreography, follow-through, transitions).
  2. 6 Render, remux, encode HyperFrames renders the video, the measured mix is remuxed back in, and an optional social encode produces a smaller file. Square or vertical editions come from scripts/reframe.py with data-ratio overrides, reusing the master's audio and timing. Posting stays with you.

Syntax, arguments, subcommands

How to invoke the skill: syntax, positional arguments, shared options, then each subcommand with its own syntax and outcome.

Syntax

/ak:motion-video [brief or project] [--poor]

Arguments

[brief or project]

Brief or project

Natural-language brief (topic and facts source, style or style mix, duration, voice language, where to stop) or an existing project to restyle, rework, or reframe. It is not a fixed parser command.

/ak:motion-video "Make a 60-second release video for our v3.4 notes in the glass-keynote style, using only facts from CHANGELOG.md"

Shared options

--poor

Lean single session

One session and no subagents, with lean reads, a compact brief, and one review round for the lowest token spend. It never skips a check.

/ak:motion-video "45-second changelog video in the glass-keynote style from CHANGELOG.md" --poor

Sample Prompt

Concrete invocations for each flag, subcommand, mode, or route available in the live workflow.

Release video, one style Recommended 
/ak:motion-video "Make a 60-second release video for our v3.4 notes in the glass-keynote style. Use only facts from CHANGELOG.md, English voice-over with Vietnamese captions, and stop after the master render; do not publish"
Use when:
Use for a launch or changelog video with one named look and a clear facts source.
Expected:
Creates a new project under assets/videos/, generates and aligns audio through multix, renders the master MP4, and reports lint, check, splice, mix, loudness, and ffprobe numbers.
Mixed style per scene 
/ak:motion-video "Apple-like base with Linear-like UI demos, blueprint diagrams, and a Nike-like finale for our feature launch"
Use when:
Use when different parts of the video need different looks on one base style.
Expected:
Resolves the mix with resolve-style.py into a per-scene table in the plan, takes each dimension from the named style, and keeps the same contract, timing, mix, and render checks.
Rework the music only 
/ak:motion-video "Re-time the drops in assets/videos/v3-4-launch to the new music and render a social encode"
Use when:
Use for a partial rework such as new music, re-timed drops, a re-mix, a re-render, or a social encode.
Expected:
Re-fits the beat grid and arrangement, reruns the timeline and remux, verifies every splice within ±2 ms, and writes a smaller social MP4 without regenerating the voice-over.
Square and vertical editions 
/ak:motion-video "Make 9:16 and 1:1 editions of assets/videos/v3-4-launch for Reels and the feed"
Use when:
Use when a finished master needs a vertical or square version without regenerating audio.
Expected:
Adds data-ratio overrides in the master, runs scripts/reframe.py for each ratio, and renders each edition after it passes the same lint, check, ffprobe frame size, loudness, and blackdetect checks.
Lean run 
/ak:motion-video "45-second changelog video in the glass-keynote style from CHANGELOG.md" --poor
Use when:
Use when token spend matters more than the full director and executor split.
Expected:
Runs the whole pipeline in one session with a compact brief and one review round, and still reports every lint, check, splice, mix, and ffprobe number.

Handled Scope

  • Pin the outcome
  • Scaffold the project
  • Generate and align audio
  • Fit beat and arrange music
  • Build timeline and composition
  • Render, remux, encode

Artifacts Produced

Motion video evidence

Pattern

Project path, style table, lint and check results, splice drift, music-under-speech dB, ffprobe streams, LUFS and peak, blackdetect result, render paths.

Done only when lint and check pass, every scene passed the director's pass, every requested 1:1 or 9:16 edition passes the same checks, every splice is within ±2 ms, music sits about 4–6 dB under speech, ffprobe shows 1920×1080 30 fps AAC 48 kHz stereo, and loudness is about −14 LUFS with peak at or below −1 dBFS.

Next