🎬 Promo Video Cheat Sheet · Day 8 · pin this tab

15 seconds of cinema.

Tool comparison · 7-shot storyboard · prompt formulas per tool · platform specs · edit checklist. By 16:00 you have a promo video ready to post.

🏆 5 golden rules · memorize before you prompt

The truths of AI video 2026.

1 · Break your video into 2-sec shots. AI models drift after 5 sec. 7 shots × 2 sec = clean 15 sec.
2 · ONE dominant motion per shot. Camera pan OR subject action · never both. Higgsfield's core rule.
3 · Dialogue in "double quotes" for lip-sync. Seedance 2.0 auto-syncs. Keep lines short.
4 · Nano Banana Pro for text. Only 2026 model reliably renders legible text-in-image. Use for thumbnails + end cards.
5 · Sound is 50% of feel. A silent AI video looks fake. Suno music bed + subtle SFX transforms perception.
🎨 The 4 formats · pick your bet

Format shapes outcome.

A · Cinematic Teaser

Movie-trailer feel. Music + text · epic visuals · 15-20 sec.
Best for: bold brand · category-defining · intrigue.
Tool: Seedance 2.0 + Nano Banana Pro titles.

B · Product Demo

Screen recording + VO. Fast · specific benefit shown · 30 sec.
Best for: SaaS · dev tools · PLG · trust.
Tool: CleanShot/Loom + Nano Banana Pro thumbnail.

C · Talking Head

You on camera. Direct · quick cuts · text overlays · 45 sec.
Best for: founder story · credibility · personal brand.
Tool: Phone camera + Descript subtitles.

D · Story-Driven

Short film. Scene → tension → reveal → CTA · 30-60 sec.
Best for: consumer app · emotional pull.
Tool: Higgsfield Cinema Studio + Motion Transfer.

🎬 The 7-shot storyboard · 15-sec structure

Each shot · 2 seconds.

#TimeShotPurposeBest tool
10:00–0:02HOOKStop the scroll · match Day 7 Founder Story hookSeedance / Higgsfield
20:02–0:04PAINConcrete moment of the painSeedance
30:04–0:06INSIGHTMoment of realization · visual metaphor OKHiggsfield Cinema Studio
40:06–0:08PRODUCT REVEALFirst glimpse · UI · logo · founder w/ phoneScreen record + Nano Banana
50:08–0:10PROOFReal users · real numbers · real screenshotsReal footage · not AI
60:10–0:12TRANSFORMATIONBefore-and-after · what world looks like afterSeedance / Higgsfield
70:12–0:14CTAProduct name · URL · one actionNano Banana Pro
End0:14–0:15END CARDLogo + URL frozen for screenshotNano Banana Pro
🌊 Seedance 2.0 · prompt formula

Native audio · lip-sync in quotes.

Formula: [SUBJECT] [ACTION] in [SETTING] · [camera movement] · [lighting] · [mood] · [physics interaction] · [text overlay if needed] · [aspect + duration].
═══ SEEDANCE 2.0 · PASTE-READY TEMPLATE ═══ [SUBJECT · e.g. "A warm-lit hand"] [ACTION · e.g. "journaling in a linen-covered notebook"] [SETTING · e.g. "on a wooden desk"] · [LIGHTING · e.g. "morning light streaming from left"] · [PHYSICS · e.g. "steam rising from a coffee cup"] · [CAMERA MOVE · e.g. "slow dolly in on the pen"]. Then cut to · [SHOT 2 description with SAME subject if consistency needed]. Then cut to · [SHOT 3 description]. Text appears reading "[YOUR HOOK LINE from Day 7]" in [brand heading font style] · [placement · centered · lower-third] · fade to logo + URL at the end. Cinematic · [color grade · e.g. "warm color grade"] · shallow depth of field · [duration · e.g. "15 seconds"] · [aspect · 9:16 · 1:1 · 16:9] · 1080p. // FOR LIP-SYNC · use quotes: [Character] looks at camera · says "[EXACT 8-word sentence]" · nods warmly · [lighting + setting].
Where to run: seedance2.ai · fal.ai/models/bytedance/seedance-v2 · or via Higgsfield's multi-model interface. Free tier: 3-5 clips/day.
🎥 Higgsfield · 5-layer prompt system

Each layer · one visual job.

Higgsfield does NOT use one text box for everything. Set all 5 layers together for stable · predictable outputs. Short direct sentences beat paragraphs.

#LayerWhat it controlsExample
1Subject + AppearanceMain subject · clothing · pose · traits"Young woman in red silk dress · confident stance · mid-30s"
2Camera + MovementONE camera move · explicit motion preset"Slow dolly-in · steady creep · rack focus"
3LightingSet in IMAGE prompt · not video (avoids flicker)"Golden hour · warm rim light from left · soft key"
4Mood + Style2-3 adjectives · color grade · film look"Cinematic · moody · teal-orange · 24fps film"
5Aspect + FramingRatio + depth of field · at end"9:16 vertical · shallow DoF · medium close-up"
═══ HIGGSFIELD · PROVEN LAYERED PROMPT FORMULA ═══ Cinematic [CAMERA MOVE] on [SUBJECT with appearance], [SETTING], camera [HOW IT MOVES], [DEPTH OF FIELD], [LIGHTING as locked property], [COLOR GRADE + FILM LOOK], [MOOD + emotional tone], [aspect + framing]. ═══ EXAMPLE ═══ Cinematic slow dolly-in on a young woman in a red silk dress standing in a fog-filled warehouse, camera glides forward at a steady creep, shallow depth of field with the background softening as we approach, volumetric god rays cutting through haze, moody teal-and-orange color grade, anamorphic lens flares, 24fps film look, dramatic and tense mood, 9:16 vertical, medium close-up.

🎬 Cinema Studio

Choose virtual lens: 35mm portrait · 85mm cinematic · 16mm wide · 200mm telephoto.

🔄 Motion Transfer

Upload reference clip · Higgsfield extracts rhythm · gestures · camera moves · applies to your gen.

🎨 Multi-Model

Kling 3.0 · Veo 3.1 · Sora 2 · Seedance 2.0 · WAN 2.7. Side-by-side output comparison.

Model picker:
ModelBest forCredits
Kling 3.0Character-driven · multi-shot · 4K · consistent voice~6 (cheapest)
Veo 3.1Cinematic scenes with audio · finished promo shots~12
Sora 2Physics realism · complex scenes~15
Seedance 2.0Native audio + lip-sync + multi-shot single prompt~15
Critical rule: ONE dominant motion per prompt. Camera pan OR subject action · never both. Violate this and every clip flickers.
⚛ Remotion · code-your-video with Cowork ⭐

The 2026 killer app for demo videos.

Seedance/Higgsfield generate FROM IMAGINATION. Remotion generates FROM YOUR ACTUAL LIVE APP. Real screenshots · real data · real trust signal. Combined with Remotion Skills (Jan 2026 · works with Cowork) · you describe the video in natural language · Cowork writes the React code · Remotion renders MP4.

What it is

React framework where every frame = a component. JSX + CSS + data → MP4 / WebM. Latest: v4.0.489. Node 16+ · macOS 15+.

🎯 Best for

Product demos · data visualizations · personalized reports · batch generation. Anything that uses YOUR real app or YOUR data.

Time to first video

~10-15 min for polished 30-sec output (research-verified). Setup: one terminal command · then everything is browser-based.

═══ REMOTION · SETUP (one-time · 5 min) ═══ # In Terminal (Cmd+Space → "Terminal"): cd ~/Desktop npx create-video@latest --yes --blank my-app-demo cd my-app-demo npm i npx remotion skills add # adds AI-agent capability npm run dev # opens localhost:3000 # Then in Cowork · File → Add Folder → ~/Desktop/my-app-demo # Cowork now sees the codebase and can generate/modify compositions.
═══ REMOTION · COWORK PROMPT FOR AUTOMATED APP DEMO ═══ Using the Remotion project in this folder · create a composition 'AppDemo' · 30 seconds · 1920×1080 · 30fps. // STEP 1: Puppeteer script to auto-capture live app screens Write scripts/capture-screens.ts using Puppeteer: - Visit [my-domain] · take full-page screenshot → public/screens/1.png - Repeat for /signup · /main-feature · /dashboard (5 total) - Also capture 60 frames of scroll animation on landing → public/scroll/ // STEP 2: Assemble into video - 0-3s: Zoom into landing · brand logo fade-in - 3-6s: Scroll animation of landing (public/scroll/001-060.png sequence) - 6-11s: Feature 1 slides in · text overlay "[Feature 1 one-liner]" - 11-16s: Feature 2 same - 16-21s: Feature 3 same - 21-26s: Dashboard pan · text "Real users · real data" - 26-30s: CTA card · "[my-domain]" · brand gradient · particle glow Match brand: primary [hex] · accent [hex]. Use spring animations + interpolate() for smooth transitions. Add public/music.mp3 as Audio component · 0.7 volume · fade out last 3s. // STEP 3: Render When done · run: npx remotion render AppDemo out/demo.mp4
The unlock: once Remotion + Cowork are wired · you can batch generate 20 personalized demo videos for 20 prospects — each with their name in the CTA · same underlying composition. Personalization at scale for Thursday's Sales Sprint.
Alternative if terminal is a blocker: use Remotion Recorder (their web-based screen-recorder tool) or Loom / CleanShot X for screen recording · then edit in CapCut. Slower but no code required.
🍌 Nano Banana Pro · thumbnail + text-in-image

The only 2026 model that renders legible text.

Formula: "Create an image of [subject] [action] [scene]." Be SPECIFIC · not "woman in dress" but "young woman in red dress running through park at sunset · shot with 85mm lens."
═══ NANO BANANA PRO · THUMBNAIL PROMPT ═══ Create a 1200×628 image for a LinkedIn video thumbnail. Product: [name from P00] Composition: [main subject] on left · large text on right reading "[HOOK LINE from Day 7 Founder Story]" in [brand heading font · e.g. bold serif italic]. Background: [brand primary color hex] gradient · subtle texture. Mood: [3 adjectives from brand tokens · e.g. "warm · precise · playful"]. Include play button icon subtly in center. Photo-realistic · high contrast · readable at mobile size.
Access: Gemini app · pick "Thinking" model · click "Create images." Google AI Plus/Pro/Ultra subscribers get higher quotas.
📱 Platform specs · export cheat table

Same video · 4 platforms.

PlatformAspectResolutionLength capCaption
LinkedIn1:1 or 16:91080×1080 · 1920×1080≤ 90 sec≤ 1300 chars
X16:91920×1080≤ 140 sec≤ 280 chars
IG Reel9:161080×1920≤ 90 sec≤ 2200 chars
TikTok9:161080×1920≤ 60 sec≤ 2200 chars
Posting schedule Wednesday: 16:00 X · 16:30 LinkedIn · 17:00 IG Reel · 17:30 TikTok · 18:00 X reply thread. Momentum matters more than perfect timing.
✂ Edit + subtitle checklist · before publish

The polish gate.

🚀 Thursday link · this is Sales Sprint ammo

The video is your ammo.

📸 Wed 16:00

Video published
4 platforms · thumbnail on LinkedIn · captions per platform

💰 Thu 12:30

Sales Sprint kickoff
Attach video to every cold email · DM · voice note. It does 3× the pitching of text alone.

🎯 Fri 14:30

Demo Day
Open your 3-min demo with the 15-sec promo · lands the room in the first 15 sec