Launch Film logo

Launch Film

Community
SkylarKitchen
launch-film

Make a short, frame-exact product film for X or a portfolio from the real thing (a web app, a site, a Three.js scene, project photos), authored as a plan and rendered on a canvas engine rather than cut from a screen recording. Use when the user wants a launch video, a teaser, a "better version" of their own screen recording, or a showcase clip of something they built. Looping background reels cut from existing footage are a different job.

Overview

PublisherSkylarKitchen
Repositoryskills
Skill namelaunch-film
Stars
132
Forks
16
Bundled files
12
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.

  • 12 bundled files

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

  • Open source

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

Installation

Install the Launch Film 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/SkylarKitchen/skills/tree/main/skills/design/launch-film
  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/SkylarKitchen/skills.git /tmp/skills
mkdir -p .claude/skills
cp -r /tmp/skills/skills/design/launch-film .claude/skills/launch-film
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Launch Film 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 Launch Film 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 Launch Film 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.

Launch film

The films are authored, not cut. Energy comes from a camera moving over one continuous surface, with eased push-ins on the details that matter, not from edits. Every frame is a pure function of time, so a render is repeatable and a 25 s film at 1080p renders in under a minute.

The engine is engine/film.html, a 2D canvas that plays a plan.js. engine/capture.cjs steps it frame by frame in headless Chromium and pipes the frames to ffmpeg.

Mode: pick one first

  • Showcase (the default): the work speaks. The opener is an accelerating flip through the project's own images (photos, drawings, stills), with no words. The body is one continuous camera over the real surface, with push-ins on the best details. It ends on the pulled-back product, or on a plain title if the user wants one. Ask before adding a title card. Add one plain caption per beat at most (people watch X muted); no taglines, no pitch copy.
  • Interaction: the cursor acts on the real page. A click motivates each camera move, and hovers show the page's real states. Shoot cursor-free stills of each state with Playwright (hover() and click() draw no cursor), plus each element's centre. Then use the product scene's states (section swaps keyed by time), hovers (a band, underline or ring where the page has no hover style), cursor.clicks and cursor.hide.
  • 3D product (a Three.js app): headless Chromium has no GPU, so render the 3D shots as frame sequences on a real GPU in a normal browser. A debug hook drives the camera and app state per frame, and each frame is POSTed as a JPEG to a local sink. Render at 4K, downscale to 1080p with lanczos, and play them in frames scenes. from: {image, wipe} wipes a photo into the render from the same camera. The UI take is the real app driven headless with real pointer events, one screenshot per frame (SwiftShader draws WebGL at about 1 s a frame), with the pointer recorded to cursor.json and replayed through a product scene with seq and cursor.frames. Dry-run each drag first and read the app's own state after it: a move that looks fine can turn a piece or break the app's checks. Worked example: examples/room-configurator/, a film of a room-from-photos page (film.js renders the shots, ui.cjs records the UI take, plan.js and plan45.js hold the timeline).

Length and rhythm

  • Teaser, 20–30 s for X: opener about 2 s, body in two or three beats of 4–8 s each.
  • Feature, about 80 s: a few continuous takes with push-ins. Calm shots have a median of 4–8 s.
  • Hard cuts only between scenes. Rapid cuts live only in the opener, never over UI states. Longer than 90 s is a tutorial, not this skill.

Pipeline

  1. Source the surface. node engine/shoot.cjs <url> page.png 1440 900 (add full for a scrolling page). For 3D, render the frame sequences first (see the worked example). Measure every position you'll push in on from the full-size image, never from a scaled contact sheet.
  2. Write plan.js in the project (copy examples/quickstart/plan.js). See the plan reference below.
  3. Render: node engine/capture.cjs plan.js preview.mp4. It needs Playwright and ffmpeg; CHROMIUM=<path> uses a system Chromium.
  4. Look at it. Make a sheet with ffmpeg -i preview.mp4 -vf "fps=2,scale=320:-1,tile=8x5" -frames:v 1 sheet.png, and pull full-size frames mid camera move. The sheet hides direction errors, such as a panel dropping in from above instead of rising.
  5. Gate. bash measure.sh preview.mp4 <opener_end_s> must pass: after the opener, calm shots have a median of at least 3.5 s, with at most two rapid bursts of up to 5 s. Then run bash holds.sh preview.mp4 <t> ... at each camera hold. It fails a push-in that frames blank page.
  6. Show the user the preview and the sheet before any final encode.
  7. Deliver. Re-render with CRF=18 PRESET=slow. X takes H.264 1920×1080. For the feed, a 4:5 cut (frame: [864, 1080], scale: 1.25 = 1080×1350) takes more of a phone screen. Give it its own plan that imports the 16:9 one and overrides the frame, the camera keys and the type sizes. A 3D film needs its shots re-rendered at 4:5: a narrower frame keeps the photo's height, not its width.

Sound

Deliver X cuts without an audio track (-an) unless the user supplies music or asks for sound: X autoplays muted, and generic synth beds read as cheap. capture.cjs writes <out>.cues.json with every cut, camera whip and press. engine/sound.py (needs numpy) turns the cues into a synth bed if the user wants one. With supplied music, snap scene durations to its beats.

Plan reference

Top level:

  • frame and scale: the logical frame (default [1280, 720]) and output scale. Positions and sizes are in logical px.
  • fps: default 25. Use 30 to match frame sequences rendered at 30, and keep every duration a whole number of frames (n / 30).
  • images: {name: url}, preloaded.
  • fonts: {Name: url} or {Name: {src, weight: "100 900"}} for a variable face; font picks the one to use. Always load one. A minimal Linux container may map sans-serif to a monospace face.
  • captions: labels in film time, {text, from, to, row, tag}. They sit top-left, in sentence case: a white label, with an optional dark letter tag such as an option's A, B or C. caption sets x, y, gap, size, small, ink and bg.
  • grain: true: adds film grain.

Scenes ({type, dur, bg, ...}), played in order:

  • still: a project image. cover: true makes it full-bleed with a slow push; otherwise fit, dx and dy place it. word sets optional text.
  • product: a UI image floating on the ground. crop: [x, y, w, h] of the image; camera: [[t, [cx, cy, zoom]], ...], eased between keys, in image px from the crop's centre (positive y looks lower on the page). seq swaps in a screenshot sequence instead of image. cursor is either {path: [[t, [x, y]]], click, clicks, hide} or {frames, fps, scale, ripple} (the pointer recorded with the take, in CSS px times scale). Also states, hovers and headline (a line over the empty ground before the UI rises in).
  • frames: a full-bleed frame sequence, seq: {src: "shot_%04d.jpg", count, fps, at}. It holds the first frame before at and the last frame after the end. from: {image, wipe: [t0, t1]} wipes it in over a still.
  • title: a centred text with an optional sub line, rising in; typeOn: true types the text instead.
  • tagline: lines rising in, one after another.

Gotchas

  • A whip (a camera jump over 400 image px between keys) adds a whoosh cue. Any other camera change stays silent.
  • The scene detector behind measure.sh misses a cut between two same-colour frames, so check a white-on-white opener by eye.
  • In holds.sh, test camera holds only. A title card is meant to read as empty.
  • Check every photo before it goes public: documents with readable text, house numbers, street signs, and names on screens.
  • The capture browser has no GPU. Keep film scenes 2D canvas, and bring 3D in as pre-rendered frames.

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 Launch Film AI skill do?

Make a short, frame-exact product film for X or a portfolio from the real thing (a web app, a site, a Three.js scene, project photos), authored as a plan and rendered on a canvas engine rather than cut from a screen recording. Use when the user wants a launch video, a teaser, a "better version" of their own screen recording, or a showcase clip of something they built. Looping background reels cut from existing footage are a different job.

Why use Launch Film on TypingMind?

Because you install it once and use it with any model. Launch Film 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 Launch Film in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/SkylarKitchen/skills/tree/main/skills/design/launch-film. 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 Launch Film?

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 Launch Film?

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

Is the Launch Film AI skill free?

Yes. It is published on GitHub by SkylarKitchen 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 👇