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.
═══ 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.
#
Layer
What it controls
Example
1
Subject + Appearance
Main subject · clothing · pose · traits
"Young woman in red silk dress · confident stance · mid-30s"
2
Camera + Movement
ONE camera move · explicit motion preset
"Slow dolly-in · steady creep · rack focus"
3
Lighting
Set in IMAGE prompt · not video (avoids flicker)
"Golden hour · warm rim light from left · soft key"
4
Mood + Style
2-3 adjectives · color grade · film look
"Cinematic · moody · teal-orange · 24fps film"
5
Aspect + Framing
Ratio + 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.
Cinematic scenes with audio · finished promo shots
~12
Sora 2
Physics realism · complex scenes
~15
Seedance 2.0
Native 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 ~/Desktopnpx create-video@latest --yes --blank my-app-democd my-app-demonpm inpx remotion skills add# adds AI-agent capabilitynpm 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.
Platform
Aspect
Resolution
Length cap
Caption
LinkedIn
1:1 or 16:9
1080×1080 · 1920×1080
≤ 90 sec
≤ 1300 chars
X
16:9
1920×1080
≤ 140 sec
≤ 280 chars
IG Reel
9:16
1080×1920
≤ 90 sec
≤ 2200 chars
TikTok
9:16
1080×1920
≤ 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.
All 7 shots assembled in CapCut or Descript in P02 order
ONE transition style used throughout (hard cut · match cut · whip pan · dissolve)
Subtitles present · bold sans-serif · 60pt on 9:16 · lower-third · drop shadow
CapCut auto-caption checked · errors fixed (usually 10% need editing)
Color grade applied · consistent across all shots · matches brand mood
Music track at -18dB · ducked 6dB under any voiceover
Music hits sync to product reveal (shot 4)
End card holds 1 full sec so viewers can screenshot URL
Exported per platform · correct aspect · not stretched
Thumbnail (Nano Banana Pro) uploaded to LinkedIn separately
4 platform captions written · not one caption copy-pasted
🚀 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