saas-product-demo-video
Produce a 20-45s SaaS product demo video in Remotion. Use whenever the user wants to make a product film, marketing reel, explainer, launch video, or LinkedIn/Instagram video of their app - including requests like "make a video of my product", "remotion video", "product demo reel", "launch film", "vertical reel", "9:16 ad", "explainer", or whenever screenshots + a soundtrack + brand colors are being assembled for a short cinematic render. Also use when the user has b-roll footage (collage, stock, live action) they want layered into a motion-graphics film, or when the product has no UI to record (API-only, Chrome extension, early-stage) and schematic UIs need to be invented.
noamdorr · 站主发布
安装与使用
运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。
npx skills add https://github.com/noamdorr/saas-product-demo-video/tree/6ddaee236ea94fc2ff6d6abfddb663fdef4fe72e/saas-product-demo-video --skill "saas-product-demo-video"技能说明
name: saas-product-demo-video description: Produce a 20-45s SaaS product demo video in Remotion. Use whenever the user wants to make a product film, marketing reel, explainer, launch video, or LinkedIn/Instagram video of their app - including requests like "make a video of my product", "remotion video", "product demo reel", "launch film", "vertical reel", "9:16 ad", "explainer", or whenever screenshots + a soundtrack + brand colors are being assembled for a short cinematic render. Also use when the user has b-roll footage (collage, stock, live action) they want layered into a motion-graphics film, or when the product has no UI to record (API-only, Chrome extension, early-stage) and schematic UIs need to be invented.
SaaS Product Demo Video
End-to-end workflow for shipping a beat-synced SaaS product film in Remotion. Covers silent cinematic films, beat-locked horizontal demos, vertical reels, and films with collage b-roll layers.
The skill is opinionated about three things: assets, vibe, beats. Skip any one and you produce a generic Canva-style motion graphics reel that looks like every other SaaS demo. The discipline is what makes the difference.
These are strong recommendations rather than hard rules. Most productions follow them; some break them with intent (e.g. voice-over-driven films where beat-sync would fight the VO). When you break a rule, do it knowingly.
Prerequisites
Before starting, verify these are in place. The first one is genuinely required; the others are strong recommendations with documented fallbacks.
-
Remotion installed (required).
@remotion/cli4.0.x +@remotion/google-fonts+ React 19. Seereferences/project-scaffold.mdfor the exact minimalpackage.json/tsconfig.json/remotion.config.ts- never runnpx create-videoinside an existing git repo (it blocks on an arrow-key prompt). -
Super Powers plugin (strongly recommended, not required). When available, this skill orchestrates
superpowers:using-superpowers,superpowers:brainstorming(with visual companion),superpowers:writing-plans, andsuperpowers:subagent-driven-development. They give you HTML mockups in the browser for layout decisions, formal plan documents, and the implementer + reviewer subagent loop.If superpowers isn't installed, the skill still works - the workflow becomes more conversational and a bit slower. Brainstorming becomes inline dialogue (no browser companion); plans are written directly in the chat; scenes are implemented inline rather than via dispatched subagents. The three pillars (assets / vibe / beats) plus b-roll, schematic UI, click positioning, writing style, and the scripts in
scripts/all work the same. Don't decline the work because of a missing plugin - adapt and proceed. -
Python 3 + librosa available. Beat detection runs
scripts/detect-beats.pywhich importslibrosa. If not installed, runpip install --break-system-packages librosa numpy soundfileand tell the user - don't do it silently. -
For reference-video analysis (Pillar 2): the skill supports three paths in priority order. Use the highest one that's available - see
references/reference-video-vibe.mdfor full detection logic.geminiCLI on PATH (official Google CLI, or community wrappers likegemini-yt). Run with the prompt + YouTube URL as args.GEMINI_API_KEYset in env or.env. Use the skill's curl wrapper atscripts/analyze-reference-video.sh.- None of the above - fall back to manual Google AI Studio copy-paste. Awkward but always works; quality is comparable.
Run references/prereq-check.md as your first move in a new session.
The three pillars
These are the high-leverage activities. Skipping one to "save time" almost always costs more time downstream.
Pillar 1 - Asset intake onboarding
Before anything else, run the structured questionnaire in assets/prompt-templates/asset-intake-questionnaire.md. It asks for:
- Aspect ratio (16:9 horizontal, 9:16 vertical, or both in parallel - see
references/vertical-port.mdif both) - Duration in seconds (20-45s recommended)
- Logo - SVG preferred. If PNG-only, capture native dimensions to prevent pixelation when scaled up.
- Brand hex colors (primary accent + 3-6 supporting pastels)
- Font stack (all must be on Google Fonts)
- Copy / script - prefer a user-supplied draft over generating one. If they have no script, see
references/script-development.md. - Voiceover (optional)
- Soundtrack file - required for beat-sync. If user has none, pick from royalty-free stock.
Reference references/asset-intake.md for the full intake flow including the partial-asset minimum-viable set, naming hygiene on copy-in, and the asset-upscaling pixelation check.
Pillar 2 - Reference-video vibe workflow
Before writing scenes, ask the user for 2-3 YouTube reference videos whose vibe they want. Then produce a director's breakdown for each.
Two paths supported (see references/reference-video-vibe.md):
- Automated (preferred when available) - if the user has a
GEMINI_API_KEYset, runscripts/analyze-reference-video.sh <youtube-url> <slug>for each video. Uses Gemini Pro Preview via curl. Takes ~30s per video. Outputs land atreferences/gemini-<slug>.md. - Manual (always works) - open
aistudio.google.com, paste the prompt fromscripts/reference-video-prompt.mdalong with the YouTube URL, copy the result back into a file. Awkward but produces equally usable output. Use this when the API isn't available.
The directorial verb vocabulary (push-through, whip-pan, V-cut pull-back, ink-mask, match-cut zoom) is the most load-bearing output - it gives you specific motion primitives to build around.
Pillar 3 - Soundtrack beat-sync
Scene boundaries should land on snares (the rhetorical "drum hit" feel). Internal events (clicks, card landings, counter starts) should land on either kicks or snares within the
阅读完整来源 ↗