Claude Skill

saas-launch-demo

SaaS launch and product demo editor for Claude: turns a screen recording plus optional founder voice/face into a 60-180 s launch or demo video (outcome, problem, walkthrough, result, CTA) with readable UI, targeted zoom-ins, callouts, cursor pointers, PII blurred, and a 15 s soci

LLM Mart · 0 points · 0 views 0 listing impressions 0 install-command copies
Download viprasol-tech-claude-video-editing-skills-plugins_video-studio_skills_saas-launch-demo-0293497.zip · 3 KB

Install

skills CLI npx skills add https://github.com/Viprasol-Tech/claude-video-editing-skills/tree/main/plugins/video-studio/skills/saas-launch-demo
Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install viprasol-tech-claude-video-editing-skills@llmmart
Git git clone https://github.com/Viprasol-Tech/claude-video-editing-skills.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole viprasol-tech/claude-video-editing-skills collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

SaaS Launch Demo

Get the viewer to the "aha" fast: the outcome first, then the three moves that make it happen.

  • Best for: product launches, feature announcements, Product Hunt videos, landing-page demos, onboarding videos
  • Inputs: screen recording(s), optional webcam/founder clip or voice-over, logo, product name and URL
  • Default skin: tutorial-steps · alternatives device-fly, terminal-tech, kinetic-launch
  • Formats: 16:9 hero, 9:16 and 1:1 social cuts, GIF/WebP loops
  • Durations: hero 60-180 s (launch film 60-90 s), social 15 s, loop 6 s
  • Genre preset: --genre saas-launch-demo

Before you start

  • Needs video-studio-core in the same skills folder. CORE = that folder, SKILLS = the skills folder.
  • First run: python CORE/scripts/check_env.py.
  • Record clean: demo account with fake data, notifications off, 1920x1080 or larger, large UI zoom (125-150%).
  • PII: real emails, names, API keys or customer data on screen must be hidden. Blur them with the plan's blurs boxes (see Refine), cover them with a shape/card, or re-record with demo data.
  • Ask for: product name spelling, the one outcome to promise, the three features to show, CTA and URL, pricing claims.

How a pro cuts this

Structure (% of length)

beat % what goes here
outcome 10 the finished result on screen in the first 2 s (dashboard, report, shipped thing)
problem 10 the painful old way in one line or one shot
walkthrough 60 three features, one idea each, each ending on its result
result 10 the payoff number or screen, said and shown
CTA 10 product name, URL, offer; logo

Pacing. ASL 3-6 s; a visual change every 4 s (zoom, callout, pointer, cut). Cut every wait: loading, typing, scrolling to find something. 2x-4x speed through boring but necessary steps.

Cut rules.

  • Cut on the click: the result appears right after the action; never watch a spinner.
  • Zoom in when the UI text is below ~24 px at 1080p; zoom out to re-orient after 2-3 moves.
  • Jump cuts are fine in the screen recording; in the face cam cover them with the screen.
  • Never show a feature that does not exist yet as if it ships (label "coming soon").

Transition grammar. Cuts and zoom punches; whip only between features; dissolve only for a scene change.

Music and sound. Light electronic bed ~-26 dB, ducked under the voice. UI SFX low: click / ui-tap on real clicks, success-chime on the result, whoosh-short on feature changes (about 4/min).

Colour. No look on the UI (grade none): brand colours must stay accurate. Face cam only: cool-corporate 0.4.

Captions. Two-line subtitles low, sentence case, <= 20 cps ("preset": "clean"); keyboard shortcuts as chips.

What the client judges. UI readability (on a phone too) and time to "aha".

Common mistakes. A 15 s logo intro; tiny unreadable UI; the cursor wandering; real customer data visible; showing every setting; a feature tour with no outcome; CTA without a URL.

Make the edit

With voice in the recording (or a face cam with the screen recording as a clip):

python CORE/scripts/run.py demo.mp4 --style SKILLS/tutorial-steps --genre saas-launch-demo --format 16:9 --smart-cuts --retakes --title "Ship reports in 5 minutes" --cta "Try it free at acme.io" --work sd --stop-after plan
# face cam + separate screen recording: the screen becomes clip "main"
python CORE/scripts/run.py founder.mp4 --style SKILLS/device-fly --genre saas-launch-demo --format 16:9 --clip screen.mp4 --logo logo.png --work sd2 --stop-after plan
python CORE/scripts/render.py sd/plan.json --frames 1,6,15,30,50
python CORE/scripts/render.py sd/plan.json
python CORE/scripts/finish.py sd/plan.json --platform youtube

Silent screen recording with a written script: make the voice-over first, then use it as the speaker audio:

python CORE/scripts/tts.py --script vo.txt --out vo/voice.wav --words
python CORE/scripts/scenes.py screen.mp4 --work sd3 --threshold 0.15     # UI changes = candidate cut points

Then cut the screen to the voice's length (a timeline spec with "audio": "mute" and "speed": 2 on waits, built with timeline.py build spec.json --work scr --no-transcribe), mux the voice on with FFmpeg (ffmpeg -i scr/assembled.mp4 -i vo/voice.wav -map 0:v -map 1:a -c:v copy -shortest demo_vo.mp4) and run the first run.py command on demo_vo.mp4. Don't put a voice in "music": it would be ducked and treated as a bed.

15 s social cut: a new work folder, --format 9:16 --duration 15, outcome + one feature + CTA only.

Refine the plan

  • Hook (first 2 s): a text element with the outcome ("Reports in 5 minutes, not 5 hours"); show the result screen.
  • Zoom into UI: "speaker": {"zooms": [{"start": 8.2, "end": 11.0, "scale": 1.6, "cx": 0.72, "cy": 0.3}]} (when the screen is the speaker video; cx/cy = the button, 0-1 of the source frame, one per zoom). For a clip element use keys ("s": [[0, 1], [0.6, 1.6, "expoOut"]]).
  • Pointers and callouts: {"type": "callout", "point": [1210, 340], "x": 1500, "y": 180, "title": "One-click export", "start": 12, "end": 15}, {"type": "pointer", "x": 1210, "y": 340, "kind": "hand", "start": 12, "end": 13.5}.
  • Steps: {"type": "steps", "steps": [{"label": "Connect", "at": 9}, {"label": "Build", "at": 21}, {"label": "Share", "at": 34}]}.
  • Shortcut chips: {"type": "chip", "text": "Ctrl + K", "start": 18.4, "end": 20}; code: codeblock with cps.
  • Hide PII: "blurs": [{"x": 640, "y": 212, "w": 380, "h": 40, "start": 0, "end": 60}] (output pixels, top-left box; "pixelate": true for a mosaic; one box per time range when it scrolls or a zoom moves it). It blurs the screen under the graphics. Screen inside a clip/device element: blurs don't reach it, so use a solid {"type": "shape", "kind": "bar", ...} over it instead.
  • Device frame: {"type": "clip", "clip": "main", "device": "laptop", "motion": "flyIn"} for the opener.
  • SFX: "sfx": [{"t": 12.1, "kind": "click", "gain": 0.35}, {"t": 15.0, "kind": "success-chime", "gain": 0.4}].
  • Colour: leave the screen untouched ("color": {"auto": false}); face cam only {"look": "cool-corporate", "intensity": 0.4}.
  • Check auto counters: value and label must match what is said.

Deliver

python CORE/scripts/deliver.py sd/plan.json --to youtube,linkedin,x,gif,webp --out-dir sd/deliver
python CORE/scripts/deliver.py sd15/plan.json --to reels,tiktok,shorts --out-dir sd15/deliver
python CORE/scripts/qc.py sd/deliver/final_youtube.mp4 --plan sd/plan.json --platform youtube --strict

gif/webp give a 6 s loop of the start for the landing page: make the first 6 s the outcome. Product Hunt and landing pages autoplay muted, so burn captions in (--burn yes) on those uploads.

Check before delivering

  • Watch the contact sheet at phone size: every UI label readable, zooms land on the right element.
  • No real personal data, keys or customer names visible in any frame (scrub the full render).
  • Product name, prices and URL correct; every promised feature exists.
  • qc.py passes; cut waits are gone; the CTA holds >= 2 s.

Example requests

  • "Turn this 6-minute screen recording into a 90-second launch video for Product Hunt."
  • "Make a 15-second vertical teaser of the new export feature with captions."
  • "Zoom into the buttons I click and blur the customer emails."
Files (claude-video-editing-skills)
  • SKILL.md 8 KB
    ---
    name: saas-launch-demo
    description: "SaaS launch and product demo editor for Claude: turns a screen recording plus optional founder voice/face into a 60-180 s launch or demo video (outcome, problem, walkthrough, result, CTA) with readable UI, targeted zoom-ins, callouts, cursor pointers, PII blurred, and a 15 s social cut and GIF loop. Use when the user says 'edit my product demo', 'make a launch video for my app', 'Product Hunt video', 'feature announcement video', 'turn this screen recording into a demo', 'SaaS explainer', or hands over a screen capture of software. Runs on video-studio-core (transcribe, cut, motion graphics, deliver, QC)."
    license: MIT
    ---
    
    # SaaS Launch Demo
    
    Get the viewer to the "aha" fast: the outcome first, then the three moves that make it happen.
    
    - **Best for:** product launches, feature announcements, Product Hunt videos, landing-page demos, onboarding videos
    - **Inputs:** screen recording(s), optional webcam/founder clip or voice-over, logo, product name and URL
    - **Default skin:** `tutorial-steps` · alternatives `device-fly`, `terminal-tech`, `kinetic-launch`
    - **Formats:** 16:9 hero, 9:16 and 1:1 social cuts, GIF/WebP loops
    - **Durations:** hero 60-180 s (launch film 60-90 s), social 15 s, loop 6 s
    - **Genre preset:** `--genre saas-launch-demo`
    
    ## Before you start
    
    - Needs **video-studio-core** in the same skills folder. `CORE` = that folder, `SKILLS` = the skills folder.
    - First run: `python CORE/scripts/check_env.py`.
    - Record clean: demo account with fake data, notifications off, 1920x1080 or larger, large UI zoom (125-150%).
    - **PII:** real emails, names, API keys or customer data on screen must be hidden. Blur them with the plan's
      `blurs` boxes (see Refine), cover them with a `shape`/`card`, or re-record with demo data.
    - Ask for: product name spelling, the one outcome to promise, the three features to show, CTA and URL, pricing claims.
    
    ## How a pro cuts this
    
    **Structure (% of length)**
    
    | beat | % | what goes here |
    | --- | --- | --- |
    | outcome | 10 | the finished result on screen in the first 2 s (dashboard, report, shipped thing) |
    | problem | 10 | the painful old way in one line or one shot |
    | walkthrough | 60 | three features, one idea each, each ending on its result |
    | result | 10 | the payoff number or screen, said and shown |
    | CTA | 10 | product name, URL, offer; logo |
    
    **Pacing.** ASL 3-6 s; a visual change every 4 s (zoom, callout, pointer, cut). Cut every wait: loading, typing,
    scrolling to find something. 2x-4x speed through boring but necessary steps.
    
    **Cut rules.**
    - Cut on the click: the result appears right after the action; never watch a spinner.
    - Zoom in when the UI text is below ~24 px at 1080p; zoom out to re-orient after 2-3 moves.
    - Jump cuts are fine in the screen recording; in the face cam cover them with the screen.
    - Never show a feature that does not exist yet as if it ships (label "coming soon").
    
    **Transition grammar.** Cuts and zoom punches; whip only between features; dissolve only for a scene change.
    
    **Music and sound.** Light electronic bed ~-26 dB, ducked under the voice. UI SFX low: `click` / `ui-tap` on real
    clicks, `success-chime` on the result, `whoosh-short` on feature changes (about 4/min).
    
    **Colour.** No look on the UI (`grade` none): brand colours must stay accurate. Face cam only: `cool-corporate` 0.4.
    
    **Captions.** Two-line subtitles low, sentence case, <= 20 cps (`"preset": "clean"`); keyboard shortcuts as `chip`s.
    
    **What the client judges.** UI readability (on a phone too) and time to "aha".
    
    **Common mistakes.** A 15 s logo intro; tiny unreadable UI; the cursor wandering; real customer data visible;
    showing every setting; a feature tour with no outcome; CTA without a URL.
    
    ## Make the edit
    
    With voice in the recording (or a face cam with the screen recording as a clip):
    
    ```bash
    python CORE/scripts/run.py demo.mp4 --style SKILLS/tutorial-steps --genre saas-launch-demo --format 16:9 --smart-cuts --retakes --title "Ship reports in 5 minutes" --cta "Try it free at acme.io" --work sd --stop-after plan
    # face cam + separate screen recording: the screen becomes clip "main"
    python CORE/scripts/run.py founder.mp4 --style SKILLS/device-fly --genre saas-launch-demo --format 16:9 --clip screen.mp4 --logo logo.png --work sd2 --stop-after plan
    python CORE/scripts/render.py sd/plan.json --frames 1,6,15,30,50
    python CORE/scripts/render.py sd/plan.json
    python CORE/scripts/finish.py sd/plan.json --platform youtube
    ```
    
    Silent screen recording with a written script: make the voice-over first, then use it as the speaker audio:
    
    ```bash
    python CORE/scripts/tts.py --script vo.txt --out vo/voice.wav --words
    python CORE/scripts/scenes.py screen.mp4 --work sd3 --threshold 0.15     # UI changes = candidate cut points
    ```
    
    Then cut the screen to the voice's length (a timeline spec with `"audio": "mute"` and `"speed": 2` on waits,
    built with `timeline.py build spec.json --work scr --no-transcribe`), mux the voice on with FFmpeg
    (`ffmpeg -i scr/assembled.mp4 -i vo/voice.wav -map 0:v -map 1:a -c:v copy -shortest demo_vo.mp4`) and run the
    first `run.py` command on `demo_vo.mp4`. Don't put a voice in `"music"`: it would be ducked and treated as a bed.
    
    **15 s social cut:** a new work folder, `--format 9:16 --duration 15`, outcome + one feature + CTA only.
    
    ## Refine the plan
    
    - **Hook (first 2 s):** a `text` element with the outcome ("Reports in 5 minutes, not 5 hours"); show the result screen.
    - **Zoom into UI:** `"speaker": {"zooms": [{"start": 8.2, "end": 11.0, "scale": 1.6, "cx": 0.72, "cy": 0.3}]}`
      (when the screen is the speaker video; `cx`/`cy` = the button, 0-1 of the source frame, one per zoom). For a `clip` element use `keys` (`"s": [[0, 1], [0.6, 1.6, "expoOut"]]`).
    - **Pointers and callouts:** `{"type": "callout", "point": [1210, 340], "x": 1500, "y": 180, "title": "One-click export", "start": 12, "end": 15}`,
      `{"type": "pointer", "x": 1210, "y": 340, "kind": "hand", "start": 12, "end": 13.5}`.
    - **Steps:** `{"type": "steps", "steps": [{"label": "Connect", "at": 9}, {"label": "Build", "at": 21}, {"label": "Share", "at": 34}]}`.
    - **Shortcut chips:** `{"type": "chip", "text": "Ctrl + K", "start": 18.4, "end": 20}`; code: `codeblock` with `cps`.
    - **Hide PII:** `"blurs": [{"x": 640, "y": 212, "w": 380, "h": 40, "start": 0, "end": 60}]` (output pixels, top-left
      box; `"pixelate": true` for a mosaic; one box per time range when it scrolls or a zoom moves it). It blurs the
      screen under the graphics. Screen inside a `clip`/device element: blurs don't reach it, so use a solid
      `{"type": "shape", "kind": "bar", ...}` over it instead.
    - **Device frame:** `{"type": "clip", "clip": "main", "device": "laptop", "motion": "flyIn"}` for the opener.
    - **SFX:** `"sfx": [{"t": 12.1, "kind": "click", "gain": 0.35}, {"t": 15.0, "kind": "success-chime", "gain": 0.4}]`.
    - **Colour:** leave the screen untouched (`"color": {"auto": false}`); face cam only `{"look": "cool-corporate", "intensity": 0.4}`.
    - Check auto `counter`s: value and label must match what is said.
    
    ## Deliver
    
    ```bash
    python CORE/scripts/deliver.py sd/plan.json --to youtube,linkedin,x,gif,webp --out-dir sd/deliver
    python CORE/scripts/deliver.py sd15/plan.json --to reels,tiktok,shorts --out-dir sd15/deliver
    python CORE/scripts/qc.py sd/deliver/final_youtube.mp4 --plan sd/plan.json --platform youtube --strict
    ```
    
    `gif`/`webp` give a 6 s loop of the start for the landing page: make the first 6 s the outcome. Product Hunt and
    landing pages autoplay muted, so burn captions in (`--burn yes`) on those uploads.
    
    ## Check before delivering
    
    - Watch the contact sheet at phone size: every UI label readable, zooms land on the right element.
    - No real personal data, keys or customer names visible in any frame (scrub the full render).
    - Product name, prices and URL correct; every promised feature exists.
    - `qc.py` passes; cut waits are gone; the CTA holds >= 2 s.
    
    ## Example requests
    
    - "Turn this 6-minute screen recording into a 90-second launch video for Product Hunt."
    - "Make a 15-second vertical teaser of the new export feature with captions."
    - "Zoom into the buttons I click and blur the customer emails."
    

Comments (0)

Sign in to join the conversation.

No comments yet.

Reviews (0)

No reviews yet.