AI video creation skill

AI-powered video creation for marketing — step-by-step workflows for HeyGen (AI spokespersons), Synthesia (AI avatars), Runway (generative video), Captions (AI editing), Opus Clip (AI clipping), ElevenLabs (AI voiceover), and Descript (AI editing).

by LeadMagic·MIT license·★ 62 Stars on the repo·GitHub ↗

Use now

Inside this skill

SKILL.md · the author's own words
# AI Video Creation

## Overview

Video is the highest-effort, highest-impact marketing format. Traditionally,
video required a camera, lighting, editing skills, and hours of post-production.
AI flips this: you can now generate professional video from text alone. AI
spokespersons deliver personalized demos. AI editors cut and caption in minutes.
Generative AI creates B-roll from prompts. This skill covers every AI video
workflow relevant to B2B GTM — from personalized outreach to social content
at scale.

## Authoritative Foundations

- **Gary Vaynerchuk** — Volume of content wins. AI makes volume possible.
- **HeyGen/Synthesia** — AI spokesperson technology

Word-for-word excerpt, not a result. Read the full SKILL.md ↓

Files of AI video creation

LeadMagic/main1 file shown
SKILL.md
Show the full text335 lines

AI Video Creation

Overview

Video is the highest-effort, highest-impact marketing format. Traditionally, video required a camera, lighting, editing skills, and hours of post-production. AI flips this: you can now generate professional video from text alone. AI spokespersons deliver personalized demos. AI editors cut and caption in minutes. Generative AI creates B-roll from prompts. This skill covers every AI video workflow relevant to B2B GTM — from personalized outreach to social content at scale.

Authoritative Foundations

  • Gary Vaynerchuk — Volume of content wins. AI makes volume possible.
  • HeyGen/Synthesia — AI spokesperson technology
  • Runway — Generative AI video
  • Opus Clip — AI-powered video clipping
  • Descript — AI-first video editing

When to Use

Trigger phrases: "AI video creation", "HeyGen tutorial", "Synthesia demo", "AI spokesperson video", "Runway for marketing", "AI video ad", "Opus Clip", "AI video editing", "ElevenLabs voiceover", "personalized video at scale", "AI video content"

The AI Video Stack

Tool What It Does Best For Cost
HeyGen AI spokesperson, avatar video, personalized video Personalized demos, outreach videos $29-89/mo
Synthesia AI avatar video, 140+ languages Training, onboarding, multi-language $29-89/mo
Runway Generative video, text-to-video, B-roll B-roll footage, visual effects $15-95/mo
Captions AI video editing, auto-captions, eye contact fix Talking-head social video $16-25/mo
Opus Clip AI clipping — long → short clips Podcast/sermon clips for social $19-29/mo
Descript AI video editor, transcript-based editing Full video production $19-29/mo
ElevenLabs AI voiceover, text-to-speech, voice cloning Narration, ads, podcasts $5-99/mo
Pika Generative video, animate images Quick video generation $10-35/mo

Step-by-Step Workflows

Workflow 1: AI Spokesperson Demo Video (1 hour)

Use case: Product demo, explainer video, "welcome" video for onboarding.

STEP 1 — Script (15 min, Claude):
"Write a 90-second product demo script for [product]. 
Audience: [ICP]. They care about [top 3 concerns].
Structure:
- 0:00-0:10: Hook — the problem they face
- 0:10-0:30: How [product] solves it (3 key features, outcomes-based)
- 0:30-0:60: Screen recording + walkthrough (you record this)
- 0:60-0:80: Social proof + results ('X teams use us, achieving Y')
- 0:80-0:90: CTA — 'Start free trial at [URL]'

Platform: spoken by AI avatar. Keep sentences short — under 15 words."

STEP 2 — Voiceover (5 min, ElevenLabs):
- Paste script into ElevenLabs Text-to-Speech
- Pick a natural voice (Adam, Bella, or similar — test 3+)
- Generate. Listen. Re-generate with pacing adjustments if needed.
- Download .mp3

STEP 3 — Avatar Video (10 min, HeyGen / Synthesia):
- Select AI avatar (matching your target audience's expectations)
- Upload script text
- Upload ElevenLabs voiceover (or use built-in TTS)
- Add: background (office, studio, or branded), on-screen text overlays
- Generate

STEP 4 — Screen Recording (15 min):
- Record product walkthrough (Loom, OBS, or QuickTime)
- Edit to match script timing

STEP 5 — Combine (15 min, Descript / CapCut):
- Layer 1: AI avatar video (talking head)
- Layer 2: Screen recording (when script references product)
- Layer 3: B-roll from Runway (generated visuals)
- Add: captions (auto-generated), background music (low, instrumental)
- Add: intro/outro branding
- Export: 1080p, 16:9
Workflow 2: Personalized Outreach Video at Scale (30 min per video)

Use case: "Hey [first name], I made this quick video for you..." Personalized videos get 3-5x higher reply rates than text-only outreach.

STEP 1 — Template Script (10 min, Claude):
"Write a 45-second personalized outreach video script. 
Personalization tokens: [first_name], [company], [pain_point].
Structure:
- 'Hey [first_name] — [relevant observation about their company/role]'
- 'The reason I'm reaching out: [pain_point is common in their industry]'
- 'We built [product] to solve exactly that. Here's what it does in 10 sec'
- 'I'd love 10 minutes to show you how it works. [Calendly] in comments'"

STEP 2 — Record Template Video (once):
- Record the template script with all personalization tokens clearly marked
- OR: Use HeyGen to generate an AI avatar video template

STEP 3 — Personalize (per prospect, 5 min):
- Research prospect: company, role, recent news, common connection
- Customize the observation: "I saw you just raised your Series A — congrats!"
- Record personalized intro (10 sec on Loom) or use HeyGen's personalization

STEP 4 — Edit & Send (5 min, Descript / Captions):
- Replace template intro with personalized intro
- Add prospect's name as on-screen text in opening frame
- Export. Upload. Send via email or LinkedIn DM.
Workflow 3: Social Short-Form Video at Scale (1 hour for 5 videos)

Use case: LinkedIn video, TikTok, YouTube Shorts, Instagram Reels.

STEP 1 — Batch Scripts (15 min, Claude):
"Write 5 short-form video scripts (60 seconds each) about [topics].
Structure:
- Hook (0-3 sec): bold claim, surprising stat, or provocative question
- Body (3-50 sec): one key insight explained with example
- CTA (50-60 sec): 'Follow for more [topic] insights'

Topics:
1. [topic]
2. [topic]
..."

STEP 2 — Batch Voiceover (10 min, ElevenLabs):
- All 5 scripts → ElevenLabs. Same voice.
- Generate as a batch.
- Download .mp3 files.

STEP 3 — Batch B-Roll (15 min, Runway):
- Generate 3-5 short clips per script (prompt: B-roll concept)
- OR: Use stock footage (Pexels, Pixabay — free)

STEP 4 — Batch Edit (20 min, Captions / Descript):
- Template: voiceover + B-roll + captions + branding
- Apply template to all 5 videos
- Auto-generate captions (Captions does this natively)
- Add: semi-transparent overlay at bottom with your handle
- Add: captions with word-by-word highlighting (95%+ watch time increase)

STEP 5 — Schedule:
- Upload all 5 to scheduling tool (Buffer, Hootsuite, native schedulers)
- Post 1/day for a workweek
Workflow 4: AI Video Ad Creative (2 hours for 5 variants)
STEP 1 — Strategy (15 min, Claude):
"Design a video ad strategy for [product] on [LinkedIn / Meta / TikTok].
Audience: [ICP]. Budget: $X. Goal: [demos / signups / leads].
Generate: 5 ad concepts with hooks, scripts, and visual direction."

STEP 2 — Scripts (15 min, Claude):
"Write 5 video ad scripts. Each: 15-30 seconds.
Variant 1: Problem/solution (hook: pain → product solves)
Variant 2: Social proof ('Join X teams...')
Variant 3: Demo teaser (show product, build curiosity)
Variant 4: Founder story ('I built this because...')
Variant 5: Objection buster ('Common belief: X. Reality: Y')"

STEP 3 — Production (60 min):
- HeyGen: Generate AI avatar delivering scripts 1, 2, 4
- Loom: Record screen for variant 3
- Runway: Generate B-roll for all variants

STEP 4 — Edit (30 min, Captions):
- Add captions (non-negotiable — 85% of social video watched on mute)
- Add branding (logo, colors, end card)
- Add background music (low, instrumental)

STEP 5 — Launch & Kill:
- Launch all 5. $50-100/day each.
- After 48 hours: kill bottom 3. Scale top 2.
- After 1 week: if CPA too high, iterate creative. Repeat.
Workflow 5: Podcast/Webinar Clips (30 min)
STEP 1 — Upload to Opus Clip or Descript:
- Upload your podcast episode (30-60 min)
- AI auto-identifies "viral moments" — high-energy, quotable segments

STEP 2 — Review & Select (5 min):
- Opus Clip suggests 5-10 clips. Pick the best 3-5.
- Each: 30-90 seconds

STEP 3 — Polish (15 min):
- Captions: auto-generated, reviewed for accuracy (names, terms)
- Add: intro frame with episode title
- Add: end card with "Full episode → [link]"
- Adjust: start/end trim points

STEP 4 — Publish (10 min):
- 1 clip/day on LinkedIn, X/Twitter, TikTok, Instagram
- Link to full episode in comments/bio

AI Video Quality Standards

What's production-ready:

  • AI spokesperson videos with real voiceover (ElevenLabs or human)
  • Screen recordings with AI narration
  • Social clips from longer content
  • AI-generated B-roll (abstract concepts, data visualization support)

What needs human touch:

  • Customer testimonials (real people > AI avatars for trust)
  • Complex product demos (AI can't anticipate all questions)
  • Brand films (emotional storytelling needs real footage)
  • Executive communications (your face > AI face for trust)

Output Format

AI VIDEO PRODUCTION PLAN

Video Type: [demo / outreach / social / ad / clip]
Platform: [LinkedIn / YouTube / TikTok / email]
Target Length: [seconds]

SCRIPT: [Claude-generated script]

PRODUCTION STACK:
- Voiceover: [ElevenLabs / human / HeyGen TTS]
- Avatar: [HeyGen / Synthesia / real human]
- B-Roll: [Runway / stock / screen recording]
- Editing: [Descript / Captions / CapCut]
- Captions: [auto / reviewed]
- Music: [source / none]

PERSONALIZATION (if applicable):
- Tokens: [first_name, company, pain_point]
- Custom intro: [recorded / AI-generated]

Implementation Checklist

  • Script under 90 seconds (attention falls off a cliff after 90s)
  • Hook in first 3 seconds (the most important 3 seconds of any video)
  • Captions on every video (85% of social video watched without sound)
  • Voiceover quality checked (natural pacing, correct pronunciation)
  • AI avatar matches audience expectations (professional for enterprise, casual for SMB)
  • Background music low enough to not compete with voice (-18dB to -24dB)
  • End card with CTA (URL, follow, subscribe)
  • File size optimized for platform (<100MB for email, <1GB for social)
  • Thumbnail generated (Canva — face + text overlay)

Quality Check

Before delivering, verify:

  • Output matches the user's stated request
  • Named frameworks or sources are reflected in the recommendation
  • The deliverable is specific enough for an agent to execute
  • Any assumptions, risks, or dependencies are explicit
  • No unsupported claims, invented facts, or private/internal references are included

Common Pitfalls

  1. AI avatar uncanny valley. Poor lip-sync, dead eyes, robotic gestures = viewers click away in 2 seconds. Fix: ElevenLabs voiceover with natural pacing. Test with 5 people: "Does this feel real?"

  2. No captions. 85% of social video is watched without sound. No captions = 85% of audience doesn't get the message. Fix: Captions on every video. Word-by-word highlighting increases watch time 95%+.

  3. Script too long. 90 seconds is the ceiling for attention on social. 15-30 seconds for ads. 45-60 seconds for demos. Fix: Cut your script in half. Then cut it again. Every word must earn its place.

  4. Generative B-roll that doesn't match. Runway generated "a team collaborating in a modern office" shows 6-fingered people at impossible desks. Fix: Use stock footage or screen recordings for concrete scenes. Use Runway for abstract concepts, backgrounds, transitions.

  5. Personalized video at scale that feels mass-produced. "Hey [first_name]" with an AI avatar that clearly isn't you. Fix: Record the personalized intro yourself (10 sec on Loom). AI handles the rest. The human touch is worth the 10 seconds.

  6. Ignoring platform specs. LinkedIn: 16:9 or 1:1, under 10 min. TikTok: 9:16 vertical, under 3 min. YouTube Shorts: 9:16, under 60 sec. Fix: Export in the correct aspect ratio and duration for each platform.

Execution Artifacts

  • references/framework-notes.md — Named frameworks and reference tables
  • templates/output-template.md — Deliverable shell for agent output
  • scripts/check-output.py — Lightweight deliverable validator
  • vibe-marketing — Full AI marketing stack
  • ai-content-creation — AI for text-based content
  • ad-creative-strategy — Creative formats per platform, testing
  • social-media-strategy — Per-platform posting and engagement
  • content-distribution — Distribution of video content across channels
  • vibe-coding — AI dev tools for building video landing pages
1---
2name: ai-video-creation
3description: >-
4 AI-powered video creation for marketing — step-by-step workflows for
5 HeyGen (AI spokespersons), Synthesia (AI avatars), Runway (generative
6 video), Captions (AI editing), Opus Clip (AI clipping), ElevenLabs
7 (AI voiceover), and Descript (AI editing). Use when creating demo
8 videos, personalized outreach videos, social video content, ads, or
9 scaling video production without a video team. Triggers on: "AI video",
10 "HeyGen", "Synthesia", "Runway AI", "AI spokesperson video", "AI ads".
11license: MIT
12compatibility: Claude Code, Codex, GitHub Copilot, Cursor, Gemini CLI, OpenCode, Goose, Hermes, Windsurf, Zed
13metadata:
14 version: "1.0.0"
15 author: LeadMagic
16 category: creative
17 tags: [ai-video, heygen, synthesia, runway, captions, elevenlabs, video-marketing, ai-avatar]
18 related_skills: [vibe-marketing, ai-content-creation, ad-creative-strategy, social-media-strategy, content-distribution, vibe-coding]
19 frameworks:
20 - "Gary Vaynerchuk — Volume of content wins. AI makes volume possible."
21 - "HeyGen/Synthesia — AI spokesperson technology"
22 - "Runway — Generative AI video"
23 - "Opus Clip — AI-powered video clipping"
24 - "Descript — AI-first video editing"
25---
26 
27# AI Video Creation
28 
29## Overview
30 
31Video is the highest-effort, highest-impact marketing format. Traditionally,
32video required a camera, lighting, editing skills, and hours of post-production.
33AI flips this: you can now generate professional video from text alone. AI
34spokespersons deliver personalized demos. AI editors cut and caption in minutes.
35Generative AI creates B-roll from prompts. This skill covers every AI video
36workflow relevant to B2B GTM — from personalized outreach to social content
37at scale.
38 
39## Authoritative Foundations
40 
41- **Gary Vaynerchuk** — Volume of content wins. AI makes volume possible.
42- **HeyGen/Synthesia** — AI spokesperson technology
43- **Runway** — Generative AI video
44- **Opus Clip** — AI-powered video clipping
45- **Descript** — AI-first video editing
46 
47## When to Use
48 
49Trigger phrases: "AI video creation", "HeyGen tutorial", "Synthesia demo",
50"AI spokesperson video", "Runway for marketing", "AI video ad", "Opus Clip",
51"AI video editing", "ElevenLabs voiceover", "personalized video at scale",
52"AI video content"
53 
54## The AI Video Stack
55 
56| Tool | What It Does | Best For | Cost |
57|---|---|---|---|
58| **HeyGen** | AI spokesperson, avatar video, personalized video | Personalized demos, outreach videos | $29-89/mo |
59| **Synthesia** | AI avatar video, 140+ languages | Training, onboarding, multi-language | $29-89/mo |
60| **Runway** | Generative video, text-to-video, B-roll | B-roll footage, visual effects | $15-95/mo |
61| **Captions** | AI video editing, auto-captions, eye contact fix | Talking-head social video | $16-25/mo |
62| **Opus Clip** | AI clipping — long → short clips | Podcast/sermon clips for social | $19-29/mo |
63| **Descript** | AI video editor, transcript-based editing | Full video production | $19-29/mo |
64| **ElevenLabs** | AI voiceover, text-to-speech, voice cloning | Narration, ads, podcasts | $5-99/mo |
65| **Pika** | Generative video, animate images | Quick video generation | $10-35/mo |
66 
67## Step-by-Step Workflows
68 
69### Workflow 1: AI Spokesperson Demo Video (1 hour)
70 
71**Use case:** Product demo, explainer video, "welcome" video for onboarding.
72 
73```
74STEP 1 — Script (15 min, Claude):
75"Write a 90-second product demo script for [product].
76Audience: [ICP]. They care about [top 3 concerns].
77Structure:
78- 0:00-0:10: Hook — the problem they face
79- 0:10-0:30: How [product] solves it (3 key features, outcomes-based)
80- 0:30-0:60: Screen recording + walkthrough (you record this)
81- 0:60-0:80: Social proof + results ('X teams use us, achieving Y')
82- 0:80-0:90: CTA — 'Start free trial at [URL]'
83 
84Platform: spoken by AI avatar. Keep sentences short — under 15 words."
85 
86STEP 2 — Voiceover (5 min, ElevenLabs):
87- Paste script into ElevenLabs Text-to-Speech
88- Pick a natural voice (Adam, Bella, or similar — test 3+)
89- Generate. Listen. Re-generate with pacing adjustments if needed.
90- Download .mp3
91 
92STEP 3 — Avatar Video (10 min, HeyGen / Synthesia):
93- Select AI avatar (matching your target audience's expectations)
94- Upload script text
95- Upload ElevenLabs voiceover (or use built-in TTS)
96- Add: background (office, studio, or branded), on-screen text overlays
97- Generate
98 
99STEP 4 — Screen Recording (15 min):
100- Record product walkthrough (Loom, OBS, or QuickTime)
101- Edit to match script timing
102 
103STEP 5 — Combine (15 min, Descript / CapCut):
104- Layer 1: AI avatar video (talking head)
105- Layer 2: Screen recording (when script references product)
106- Layer 3: B-roll from Runway (generated visuals)
107- Add: captions (auto-generated), background music (low, instrumental)
108- Add: intro/outro branding
109- Export: 1080p, 16:9
110```
111 
112### Workflow 2: Personalized Outreach Video at Scale (30 min per video)
113 
114**Use case:** "Hey [first name], I made this quick video for you..."
115Personalized videos get 3-5x higher reply rates than text-only outreach.
116 
117```
118STEP 1 — Template Script (10 min, Claude):
119"Write a 45-second personalized outreach video script.
120Personalization tokens: [first_name], [company], [pain_point].
121Structure:
122- 'Hey [first_name] — [relevant observation about their company/role]'
123- 'The reason I'm reaching out: [pain_point is common in their industry]'
124- 'We built [product] to solve exactly that. Here's what it does in 10 sec'
125- 'I'd love 10 minutes to show you how it works. [Calendly] in comments'"
126 
127STEP 2 — Record Template Video (once):
128- Record the template script with all personalization tokens clearly marked
129- OR: Use HeyGen to generate an AI avatar video template
130 
131STEP 3 — Personalize (per prospect, 5 min):
132- Research prospect: company, role, recent news, common connection
133- Customize the observation: "I saw you just raised your Series A — congrats!"
134- Record personalized intro (10 sec on Loom) or use HeyGen's personalization
135 
136STEP 4 — Edit & Send (5 min, Descript / Captions):
137- Replace template intro with personalized intro
138- Add prospect's name as on-screen text in opening frame
139- Export. Upload. Send via email or LinkedIn DM.
140```
141 
142### Workflow 3: Social Short-Form Video at Scale (1 hour for 5 videos)
143 
144**Use case:** LinkedIn video, TikTok, YouTube Shorts, Instagram Reels.
145 
146```
147STEP 1 — Batch Scripts (15 min, Claude):
148"Write 5 short-form video scripts (60 seconds each) about [topics].
149Structure:
150- Hook (0-3 sec): bold claim, surprising stat, or provocative question
151- Body (3-50 sec): one key insight explained with example
152- CTA (50-60 sec): 'Follow for more [topic] insights'
153 
154Topics:
1551. [topic]
1562. [topic]
157..."
158 
159STEP 2 — Batch Voiceover (10 min, ElevenLabs):
160- All 5 scripts → ElevenLabs. Same voice.
161- Generate as a batch.
162- Download .mp3 files.
163 
164STEP 3 — Batch B-Roll (15 min, Runway):
165- Generate 3-5 short clips per script (prompt: B-roll concept)
166- OR: Use stock footage (Pexels, Pixabay — free)
167 
168STEP 4 — Batch Edit (20 min, Captions / Descript):
169- Template: voiceover + B-roll + captions + branding
170- Apply template to all 5 videos
171- Auto-generate captions (Captions does this natively)
172- Add: semi-transparent overlay at bottom with your handle
173- Add: captions with word-by-word highlighting (95%+ watch time increase)
174 
175STEP 5 — Schedule:
176- Upload all 5 to scheduling tool (Buffer, Hootsuite, native schedulers)
177- Post 1/day for a workweek
178```
179 
180### Workflow 4: AI Video Ad Creative (2 hours for 5 variants)
181 
182```
183STEP 1 — Strategy (15 min, Claude):
184"Design a video ad strategy for [product] on [LinkedIn / Meta / TikTok].
185Audience: [ICP]. Budget: $X. Goal: [demos / signups / leads].
186Generate: 5 ad concepts with hooks, scripts, and visual direction."
187 
188STEP 2 — Scripts (15 min, Claude):
189"Write 5 video ad scripts. Each: 15-30 seconds.
190Variant 1: Problem/solution (hook: pain → product solves)
191Variant 2: Social proof ('Join X teams...')
192Variant 3: Demo teaser (show product, build curiosity)
193Variant 4: Founder story ('I built this because...')
194Variant 5: Objection buster ('Common belief: X. Reality: Y')"
195 
196STEP 3 — Production (60 min):
197- HeyGen: Generate AI avatar delivering scripts 1, 2, 4
198- Loom: Record screen for variant 3
199- Runway: Generate B-roll for all variants
200 
201STEP 4 — Edit (30 min, Captions):
202- Add captions (non-negotiable — 85% of social video watched on mute)
203- Add branding (logo, colors, end card)
204- Add background music (low, instrumental)
205 
206STEP 5 — Launch & Kill:
207- Launch all 5. $50-100/day each.
208- After 48 hours: kill bottom 3. Scale top 2.
209- After 1 week: if CPA too high, iterate creative. Repeat.
210```
211 
212### Workflow 5: Podcast/Webinar Clips (30 min)
213 
214```
215STEP 1 — Upload to Opus Clip or Descript:
216- Upload your podcast episode (30-60 min)
217- AI auto-identifies "viral moments" — high-energy, quotable segments
218 
219STEP 2 — Review & Select (5 min):
220- Opus Clip suggests 5-10 clips. Pick the best 3-5.
221- Each: 30-90 seconds
222 
223STEP 3 — Polish (15 min):
224- Captions: auto-generated, reviewed for accuracy (names, terms)
225- Add: intro frame with episode title
226- Add: end card with "Full episode → [link]"
227- Adjust: start/end trim points
228 
229STEP 4 — Publish (10 min):
230- 1 clip/day on LinkedIn, X/Twitter, TikTok, Instagram
231- Link to full episode in comments/bio
232```
233 
234## AI Video Quality Standards
235 
236**What's production-ready:**
237- AI spokesperson videos with real voiceover (ElevenLabs or human)
238- Screen recordings with AI narration
239- Social clips from longer content
240- AI-generated B-roll (abstract concepts, data visualization support)
241 
242**What needs human touch:**
243- Customer testimonials (real people > AI avatars for trust)
244- Complex product demos (AI can't anticipate all questions)
245- Brand films (emotional storytelling needs real footage)
246- Executive communications (your face > AI face for trust)
247 
248## Output Format
249 
250```
251AI VIDEO PRODUCTION PLAN
252 
253Video Type: [demo / outreach / social / ad / clip]
254Platform: [LinkedIn / YouTube / TikTok / email]
255Target Length: [seconds]
256 
257SCRIPT: [Claude-generated script]
258 
259PRODUCTION STACK:
260- Voiceover: [ElevenLabs / human / HeyGen TTS]
261- Avatar: [HeyGen / Synthesia / real human]
262- B-Roll: [Runway / stock / screen recording]
263- Editing: [Descript / Captions / CapCut]
264- Captions: [auto / reviewed]
265- Music: [source / none]
266 
267PERSONALIZATION (if applicable):
268- Tokens: [first_name, company, pain_point]
269- Custom intro: [recorded / AI-generated]
270```
271 
272## Implementation Checklist
273 
274- [ ] Script under 90 seconds (attention falls off a cliff after 90s)
275- [ ] Hook in first 3 seconds (the most important 3 seconds of any video)
276- [ ] Captions on every video (85% of social video watched without sound)
277- [ ] Voiceover quality checked (natural pacing, correct pronunciation)
278- [ ] AI avatar matches audience expectations (professional for enterprise, casual for SMB)
279- [ ] Background music low enough to not compete with voice (-18dB to -24dB)
280- [ ] End card with CTA (URL, follow, subscribe)
281- [ ] File size optimized for platform (<100MB for email, <1GB for social)
282- [ ] Thumbnail generated (Canva — face + text overlay)
283 
284## Quality Check
285 
286Before delivering, verify:
287 
288- [ ] Output matches the user's stated request
289- [ ] Named frameworks or sources are reflected in the recommendation
290- [ ] The deliverable is specific enough for an agent to execute
291- [ ] Any assumptions, risks, or dependencies are explicit
292- [ ] No unsupported claims, invented facts, or private/internal references are included
293 
294## Common Pitfalls
295 
2961. **AI avatar uncanny valley.** Poor lip-sync, dead eyes, robotic gestures =
297 viewers click away in 2 seconds. Fix: ElevenLabs voiceover with natural
298 pacing. Test with 5 people: "Does this feel real?"
299 
3002. **No captions.** 85% of social video is watched without sound. No captions
301 = 85% of audience doesn't get the message. Fix: Captions on every video.
302 Word-by-word highlighting increases watch time 95%+.
303 
3043. **Script too long.** 90 seconds is the ceiling for attention on social.
305 15-30 seconds for ads. 45-60 seconds for demos. Fix: Cut your script in
306 half. Then cut it again. Every word must earn its place.
307 
3084. **Generative B-roll that doesn't match.** Runway generated "a team
309 collaborating in a modern office" shows 6-fingered people at impossible
310 desks. Fix: Use stock footage or screen recordings for concrete scenes.
311 Use Runway for abstract concepts, backgrounds, transitions.
312 
3135. **Personalized video at scale that feels mass-produced.** "Hey [first_name]"
314 with an AI avatar that clearly isn't you. Fix: Record the personalized
315 intro yourself (10 sec on Loom). AI handles the rest. The human touch
316 is worth the 10 seconds.
317 
3186. **Ignoring platform specs.** LinkedIn: 16:9 or 1:1, under 10 min. TikTok:
319 9:16 vertical, under 3 min. YouTube Shorts: 9:16, under 60 sec. Fix:
320 Export in the correct aspect ratio and duration for each platform.
321 
322## Execution Artifacts
323 
324- `references/framework-notes.md` — Named frameworks and reference tables
325- `templates/output-template.md` — Deliverable shell for agent output
326- `scripts/check-output.py` — Lightweight deliverable validator
327 
328## Related Skills
329 
330- `vibe-marketing` — Full AI marketing stack
331- `ai-content-creation` — AI for text-based content
332- `ad-creative-strategy` — Creative formats per platform, testing
333- `social-media-strategy` — Per-platform posting and engagement
334- `content-distribution` — Distribution of video content across channels
335- `vibe-coding` — AI dev tools for building video landing pages

Discussion

Alternatives

Broll overlayOverlays silent b-roll cutaway clips onto a finished base video — the base audio keeps running underneath while the picture cuts away and returns, and the final duration always equals the base duration. Use when the user asks to add b-roll, add cutaways, or overlay clips on a video: "add b-roll to this", cut away to a product shot here", "overlay these clips on my UGC ad". Entered from an approved base video plus its transcript, which comes from POST /v1/transcripts and needs no local install; it never generates the base itself and it never extends a video.Creator · MITHiggsfield Social Hook SkillCrafts viral social media hook video prompts for TikTok, Instagram Reels, and YouTube Shorts via Higgsfield. Routes to kling3_0 (primary) or grok_video (fallback). Use when the user wants scroll-stopping hooks, viral short-form video, attention-grabbing openers, TikTok content, Reels content, Shorts content, or any social-media-optimized video.Creator · MITGab, o editor (/ogab)Gab, o editor de vídeo com IA (comando /ogab). Skill PRINCIPAL para qualquer trabalho de vídeo: criar um vídeo novo (Reels, YouTube Shorts/longo, Meta Ads, anúncio, institucional, vídeo de cliente) OU editar um vídeo/projeto/timeline no DaVinci Resolve (cortar, montar, legendar, colocar música, takes, lettering, motions, cor, áudio, exportar). Leva do roteiro até a timeline pronta no DaVinci: roteiro para aprovação, voz por IA, música, efeitos sonoros, takes de banco ou gerados, lettering animado, motions e gráficos com HyperFrames, legenda com métrica e conferência automática. Use SEMPRE que o usuário pedir para criar/fazer/editar um vídeo, mexer no DaVinci, reels, anúncio, pauta ou roteiro, mesmo sem citar a skill; as outras skills de vídeo (hyperframes, motion-graphics etc.) entram como ferramentas DENTRO deste fluxo. Comando: /ogab <pedido> ou /ogab <subcomando> (ajuda, config, update, pauta, cliente). Também cuida do Instagram: legenda do post, carrossel pronto em imagem (PNG/PDF na marca do cliente), gancho de Reel viral, hashtags, tirar cara de IA do texto, semana de conteúdo, perfil e dados do nicho.Creator · MITMuapi AI clippingTurn a long video into N viral-ready short clips with a single managed API call. Wraps muapi.ai's `/ai-clipping` endpoint, which handles transcription, highlight ranking through a virality framework (hook / emotional peak / opinion bomb / revelation / conflict / quotable / story peak / practical value), overlap dedupe, and vertical face-tracking auto-crop server-side. No local Whisper, no local LLM, no GPU.Creator · MIT