| name | software-demo-video |
| description | Create premium software demo and product videos that convert. Covers every phase from thumbnail and script through production, narration, transitions, sound design, and distribution. Use when planning, scripting, storyboarding, or reviewing any software marketing video — explainers, product tours, feature highlights, launch announcements, customer stories, or comparison videos. |
Software Demo Video
Create software product videos that look premium and actually convert. This skill covers the full pipeline: strategy, scripting, production, and distribution.
Quick Start
- Choose video type - Match to funnel stage and goal (see Video Types below)
- Write the script - Pick a framework, write to word count, read aloud (see references/script-templates.md)
- Plan production - Storyboard, select music, prepare environment (see references/production-techniques.md)
- Produce - Record, film, animate, edit, mix
- Optimize for distribution - Thumbnail, title, SEO, embedding strategy (see references/thumbnails-and-distribution.md)
Core Principles
These separate great software videos from forgettable ones. Derived from studying Sandwich Video, Apple, Stripe, Linear, Arc, Figma, Descript, and the top SaaS video agencies.
1. Start with the viewer's problem, not your features
Every top-performing software video opens with pain the viewer already feels. The product appears as the resolution, not the subject. Slack opens with communication chaos. Salesforce opens with slumping sales. Monday.com opens with "What does your ideal work look like?" Feature lists don't create desire; felt problems do.
2. One big idea per video
If the viewer forgets everything else, what must they remember? That's your video. Not three features, not five benefits. One idea.
WRONG: "Our platform has project management, time tracking,
invoicing, team chat, and file storage."
RIGHT: "Stop switching between apps. Everything your team needs
is in one place."
3. Show the magic
Sandwich Video's philosophy: identify the moments that make people go "oh, that's cool" and build the video around those moments. Find the 2-3 interactions in your product that create genuine delight and give them the most screen time.
4. Hook in 5 seconds or lose them
55% of viewers leave within 60 seconds. Your first 5 seconds must earn the next 55. Never open with a logo animation, "welcome back," or context-setting. Open with the problem, the result, or a pattern interrupt. Cold opens outperform brand intros in every retention study.
5. Sound design is the most underrated technique
Figma's click sounds, Apple's foley, Arc's deliberate silence -- audio transforms how viewers perceive visual content. Add subtle sounds to UI interactions. Sync cuts to music beats. A video with good sound design feels 2x more polished than one without.
6. Match video style to product personality
Linear's minimalism works because the product is minimal. Arc's bedroom vlogs work because the product is personal. Descript's comedy sketches work because the audience is creative. Style mismatch is worse than low production value.
7. Screen recordings don't have to look like screen recordings
Every top company has a strategy: Apple re-creates UI in motion graphics. Figma adds click sounds and cursor labels. Slack builds custom idealized screenshots. Linear's product is beautiful enough to capture raw. Stripe treats code as art. Find your approach.
Video Types
| Type | Length | Funnel | Goal | Audience |
|---|
| Explainer | 60-90s | Top | "What is it?" | Problem-aware prospects |
| Product Tour | 2-4 min | Middle | "How does it work?" | Evaluating prospects |
| Feature Highlight | 30-90s | Middle | "What can this do?" | Users + evaluators |
| Customer Story | 60-120s | Bottom | "Who else uses it?" | Decision-stage buyers |
| Launch Video | 30-90s | All | "What's new?" | Users + market |
| Comparison | 2-4 min | Bottom | "Why this over X?" | Active evaluators |
| Quick Tip | 30-90s | Post-sale | "How do I do this?" | Current users |
| Brand/Culture | 60-180s | Top | "Who are these people?" | Talent + curious |
See references/script-templates.md for detailed script structures and fill-in-the-blank templates for each type.
Script Frameworks
Pick the framework that matches your video's job. Each has a detailed template in references/script-templates.md.
Problem-Agitate-Solution (PAS) -- Most common
| Phase | Time | Words | Purpose |
|---|
| Problem | 15-20s | 30-40 | Name the pain |
| Agitate | 15-20s | 30-40 | Show consequences of inaction |
| Solution | 25-30s | 50-60 | Introduce product as resolution |
| CTA | 5-10s | 10-20 | Clear next step |
Videos using problem-solving structure show 60% higher engagement.
Before-After-Bridge (BAB) -- Best for visual products
Show transformation: the viewer's world today, their world with your product, how the product bridges the gap. Best for dashboards, design tools, analytics.
Tell-Show-Tell -- Best for tours and tutorials
Set context (tell), demonstrate live (show), reinforce value (tell). Increases retention by up to 65%. Break complex products into multiple Tell-Show-Tell modules.
Hero's Journey -- Best for customer stories
Your customer is the hero. Your product is the guide. Three acts: ordinary world (the problem), the journey (discovering and using the product), the return (transformed outcomes with metrics).
Other frameworks
- JTBD Narrative: "When [situation], I want to [job], so that [outcome]"
- Hook-Story-Offer: 5s hook, 30s narrative, 10s CTA (great for short-form)
- AIDA: Attention, Interest, Desire, Action
- Golden Circle: Why, How, What (Sinek)
- Pixar Story Spine: "Once upon a time... Every day... One day... Because of that..."
Opening Hooks
You have 5 seconds. Pick one:
| Hook Type | Example | Best For |
|---|
| Problem statement | "If you've spent 3 hours debugging deployment configs..." | Tutorials, pain-aware audiences |
| Result reveal | Show the finished output first, then explain how | Product tours, feature highlights |
| Pattern interrupt | Rapid zoom, sudden visual shift, freeze-frame + text | Social, crowded feeds |
| Cold open tease | Start at the climax: "...and that's when the build broke" | Narrative videos |
| Question | "What if your deploys never failed?" | B2B, specific segments |
| Shocking stat | "90% of support tickets are already answered in your docs" | Educational, data-driven audiences |
| Direct value | "Here are 5 CSS tricks you're not using" | Bottom-funnel, action-oriented |
Pattern interrupts in the first 5 seconds show 23% higher retention.
Cold opens beat brand intros. Cut logo animations. Hook first, brand second.
Narration
Pacing
| Video Type | WPM | Notes |
|---|
| Explainer | 130-150 | Sweet spot for comprehension |
| Tutorial | 120-130 | Slower for complex info |
| Promo/launch | 150-160 | Faster for energy |
Word count reference: 30s = ~75 words, 60s = ~150 words, 90s = ~225 words, 2min = ~300 words.
Point of View
Use second person ("you") for narration -- it draws viewers in and makes them the protagonist. Switch to first person for CTAs ("Start my free trial" converts 90% better than "Start your free trial").
Tone
- Conversational: Short sentences, contractions, direct. Best for explainers, startups.
- Professional: Measured, authoritative. Best for enterprise, security, finance.
- Enthusiastic: Energetic, forward-looking. Best for launches.
- Empathetic: Warm, understanding. Best for problem-focused intros.
Write as you speak. Read the script aloud. If you stumble, rewrite.
AI vs. Human Voice
AI voice is 60-86% cheaper and instantly updatable. But: AI narration shows 70% lower retention and 35% greater drop-off within 45 seconds. Use human voice for high-stakes content. AI is acceptable for high-volume internal/tutorial content.
ElevenLabs with-Timestamps API
Use the /v1/text-to-speech/{voice_id}/with-timestamps endpoint for word-level sync data. It returns identical audio plus character-level alignment, which you aggregate to word-level and sentence-level timing. This enables automated composition timing -- section durations are derived from measured audio, not guessed from word counts.
Per-Section Audio
Always generate one audio file per script section rather than a single concatenated narration. Concatenated files insert variable inter-section pauses (2-4 seconds each) that accumulate into 20-35 seconds of drift over 10 sections. Per-section audio, each pinned to its composition frame, eliminates cumulative drift entirely. Keep a concatenated file for preview/listening only.
Voice Clones
If a cloned voice is available (e.g., via ElevenLabs voice cloning), use it for authentic founder/developer narration at AI speed and cost. Clone quality is high enough for production use. The --voice <voice_id> flag in the narration script selects the clone.
Speed and Pacing Control
ElevenLabs voice settings that affect pacing:
stability: 0.5 -- natural variation without wanderingsimilarity_boost: 0.75 -- voice consistency across sectionsstyle: 0.3 -- subtle expressiveness without over-actinguse_speaker_boost: true -- clarity enhancement
For faster pacing, increase stability toward 0.7 and reduce style toward 0.1. For warmer, more conversational delivery, decrease stability toward 0.3 and increase style toward 0.5.
Voice Casting
| Brand Personality | Voice Style |
|---|
| Innovative/tech | Clear, confident, medium pace |
| Warm/approachable | Friendly, smile-in-the-voice |
| Enterprise | Deep, measured, baritone |
| Playful/creative | Energetic, expressive, varied inflection |
| Bold/disruptive | Fast-paced, confident, provocative |
Production Quick Reference
Full details in references/production-techniques.md.
Screen Recording
- Record at 4K/60fps, deliver at 1080p or 4K
- Move cursor at 50% natural speed, pause 0.5-1s before clicking
- Close all notifications, use clean browser profile, pre-load all screens
- Screen Studio ($89) is the current gold standard for polished demos
- Post-process: zoom 150-300% for callouts, ease all zooms (cubic/quintic), add subtle shadows
Transitions
- Hard cuts for 90%+ of transitions
- Zoom transitions for feature-to-feature navigation (0.3-0.8s)
- Dissolves only for time passage (0.5-2s)
- UI masking for branded polish (sidebar expands to cover frame, reveals new scene)
- Never use transitions as decoration. Every transition should serve the narrative.
Music
| Video Type | Mood | Tempo |
|---|
| Launch/sizzle | Upbeat, building | 120-140 BPM |
| Walkthrough | Light, unobtrusive | 90-110 BPM |
| Enterprise | Professional, warm | 80-100 BPM |
| Consumer app | Fun, modern | 110-130 BPM |
Beat-sync your cuts. Major transitions on downbeats, minor transitions on off-beats. This creates the "Apple feel" -- every cut landing on a musical beat.
Music platforms: Epidemic Sound ($10-20/mo, best for volume), Artlist ($10/mo, broad assets), Musicbed (premium, cinematic).
Narration Sync (Remotion Pipeline)
For Remotion-based productions, use the automated sync pipeline:
script.md --> narrate-sync.ts --> sync-manifest.json --> composition S constants --> audit-sync.ts --> render
Key principles:
- Per-section audio: One MP3 per script section, each pinned to its exact composition frame. Never use a single concatenated file as the composition audio source.
- Section timing formula:
ceil(narration_seconds * fps) + 15 frames padding. Always derive from measured audio, never from word count estimates.
- Sync manifest: JSON file with word-level and sentence-level timing data (start/end in both seconds and frames). Generated automatically by the narrate-sync script.
- Audit before render: Run audit-sync to detect dead frames, narration overruns, and visual gaps. Fix any issues before rendering.
Full details in references/remotion-production-pipeline.md.
Sound Effects
Add subtle clicks, whooshes, and pops to key UI interactions. Keep SFX -20 to -30 dB below voiceover. Not every click needs a sound -- reserve for meaningful moments.
Voiceover Recording
- Minimum: Blue Yeti / AT2020 USB+ ($50-130)
- Professional: Shure SM7B / RE20 ($300-450)
- Record in a quiet, treated room. Smile while recording.
- Post-processing chain: noise reduction > EQ > compression > de-essing > limiter > normalize to -16 LUFS
Thumbnail and Distribution
Full details in references/thumbnails-and-distribution.md.
Thumbnail Checklist
- 1280x720px, under 2MB, 16:9
- One clear focal point
- 3-5 words max, high contrast, readable at small sizes
- Does NOT repeat the video title
- Creates curiosity or emotional response
- Consistent with channel visual brand
- 2-3 variants ready for A/B testing
Yellow+black thumbnails yield the highest CTR at 8.2%. High-contrast thumbnails improve CTR by 20-40%.
Title Formulas
Optimal: 55-70 characters, 10-11 words. Numbers boost CTR by up to 36%.
| Pattern | Example |
|---|
| Transformation | "How [Tool] Cut Our Deploy Time by 80%" |
| Unexpected comparison | "[Tool A] vs [Tool B] -- The Winner Surprised Me" |
| Cautionary | "Stop Using [Practice] -- Do This Instead" |
| Speed/result | "Build a [Thing] in [Time] with [Tool]" |
| Discovery | "I Found the Best [Category] Tool (It's Not [Expected])" |
YouTube vs. Website
- YouTube for discovery and audience building (optimize thumbnails/titles)
- Self-hosted (Wistia/Mux) for conversion pages (YouTube embeds add 1.3-2.6MB, show competitor ads, and degrade LCP from 1.8s to 4.5s)
A/B test thumbnails. Median CTR uplift from testing: 33%.
What the Best Companies Do
Detailed analysis of 12+ companies in references/company-examples.md.
| Company | Signature Technique | Lesson |
|---|
| Apple | Beat-synced editing, macro close-ups, foley sound | Music sync creates emotion that bypasses rational evaluation |
| Stripe | Code as art, animation hierarchy, CSS-rendered devices | Developer tools can be visually elegant without dumbing down |
| Linear | Product IS the visual -- minimal post-production needed | If your product is beautiful, the simplest video treatment wins |
| Figma | Animated multi-cursors, UI click sounds | Make invisible capabilities (collaboration) physically visible |
| Arc | CEO in bedroom, YouTuber energy, $500 budget | Authenticity beats production value for shareability |
| Descript | Comedy sketch format, fourth-wall breaks | Match creative ambition to audience expectations |
| Slack | Problem-first, mockumentary humor, custom screenshots | Start with viewer's pain, not your features |
| Notion | Scenario-based narratives, named animation system | Never show a blank canvas -- always start with a specific story |
| Superhuman | Restraint as technique, static over video | Sometimes the most effective strategy is barely making videos |
| Loom | Product IS the demo (meta-demonstration) | If your product creates the content, use it |
Key Statistics
| Metric | Value | Source |
|---|
| Landing page conversion boost from video | +86% | EyeView |
| Viewers convinced to buy after watching video | 87% | Wyzowl 2026 |
| Viewers who watched an explainer before buying | 98% | Wyzowl 2026 |
| Conversion lift from A/B testing thumbnails | 33% median | YouTube data |
| First-person CTA conversion improvement | +90% | Journal of Interactive Marketing |
| Mid-roll CTA average conversion rate | ~16% | Wistia |
| AI narration retention drop vs. human | -70% | Industry data |
| Retention for educational/how-to content | 42.1% | YouTube analytics |
| Pattern interrupt effect on first-5s retention | +23% | YouTube data |
| Viewers lost in first 60 seconds | 55% | YouTube analytics |
Production Budget Tiers
| Tier | Cost | What You Get | Examples |
|---|
| DIY/authentic | $0-5K | Webcam + screen recording, personality carries it | Arc, Loom |
| Professional screen demo | $5-20K | Clean capture, motion graphics, licensed music | Raycast, Linear |
| Polished animated | $20-55K | Custom animation, pro voiceover, sound design | Slack, Monday |
| Premium production | $55-100K+ | Live action, actors, scripted narrative | Salesforce, Descript, Island |
| Elite | $250K-400K | Full agency, film-quality, custom scores | Apple, Sandwich Video |
Sweet spot for growth-stage SaaS: $8K-$25K agency range delivers strategy + script + custom art + voiceover + sound design.
Remotion Production
For programmatic video production using Remotion (React-based video framework), see references/remotion-production-pipeline.md.
Key Patterns
- Per-section audio architecture: Generate individual audio files per script section, each pinned to their exact composition frame. This eliminates cumulative drift from inter-section pauses in concatenated audio files.
- Sync pipeline:
script.md -> narrate-sync.ts -> sync-manifest.json -> composition S constants -> audit-sync.ts -> render. Fully automated from script changes through to timing.
- Section timing:
ceil(narration_seconds * fps) + 15 frames padding. Never guess. Always derive from measured audio.
- Audit tool: Catches dead frames, narration overruns, and visual gaps before rendering. Run before every render.
Effective Components
| Component | Purpose |
|---|
| ZoomPan | Subtle Ken Burns on screenshots (scale 1.0 to 1.03, never more) |
| FeatureCallout | Glassmorphism cards for feature highlights |
| SplitReveal | Before/after comparisons with animated divider |
| SceneTransition | Four types: morph, iris, slide-reveal, scale-fade |
| TypingAnimation | Simulated user input (chat, code, terminal presets) |
| DifferentiatorsStack | Numbered feature lists with spring animations |
| BrowserFrame | macOS browser chrome wrapper with traffic light dots and URL bar |
Common Bugs
- Double zoom: Do not use ZoomPan wrapper AND BrowserFrame
animated={true} simultaneously
- CTA-outro overlap: Crossfade should be max 30 frames
- TeaseCard brightness: Keep at 0.7+, never add more than 0.5px blur to preview images
- Concatenated audio drift: Never use
narration-full.mp3 as the composition audio source
Agent Review Pipeline
For structured quality review using independent reviewer agents, see references/agent-review-pipeline.md.
Six Review Dimensions
- Script Quality -- hook, pacing, visual direction, CTA, tone
- Visual Design -- consistency, animation, hierarchy, transitions, component polish
- Pacing & Timing -- script-audio alignment, audio-visual sync, flow, silent beats
- Technical Quality -- code quality, reusability, architecture, pipeline, performance
- Production Pipeline -- documentation, automation, iteration workflow
- Thumbnail -- visual impact, readability, click-worthiness, professionalism
Iteration Loop
script review --> apply feedback --> generate narration --> update composition
--> audit sync --> render --> multi-dimensional review --> iterate until all >= 7
Key Principles
- Self-assessment does not count. Use independent reviewer agents -- the builder has context blindness.
- Score progression tracking. Each review compares to previous version scores.
- Maximum two revision passes per dimension. Diminishing returns. Ship when all dimensions score >= 7.
- Gemini for thumbnails. Use the Gemini vision API for structured thumbnail scoring (readability, contrast, click-worthiness).
Anti-Patterns
Avoid these common mistakes:
- Opening with logo animation -- viewers click away before content starts
- Feature laundry list -- choose 2-3 "magic" moments instead
- Raw, unedited screen recordings -- at minimum add zoom, cursor polish, and music
- Corporate jargon in narration -- write as one human explaining to another
- Saving the product for the end -- show it in the first 15 seconds
- Same video for all audiences -- make role-specific variants for enterprise
- Ignoring sound design -- silence feels broken, bad music feels cheap
- Hardcoding UI screenshots -- design modularly so product updates don't require reshoots
- YouTube embed on conversion pages -- competitors' ads will play after your video ends
- Skipping the script -- unscripted demos ramble and lose viewers
Workflow Checklist
Pre-Production
Production
Post-Production
Distribution
