# AI Character Swap & Video Automation Skills.md

## Master the Complete AI Video Production Workflow

This guide walks you through the entire process of creating professional AI-powered video content, from character swaps to automated lead capture.

---

## Table of Contents

1. [Character Swapping with Seedance 2.5](#character-swapping)
2. [Voice Cloning & Synthesis](#voice-cloning)
3. [Audio Production & Mixing](#audio-production)
4. [Video Composition & Layout](#video-composition)
5. [Caption Generation & Sync](#caption-generation)
6. [Landing Page Lead Capture](#landing-page-setup)
7. [Webhook Flows & Automation](#webhook-flows)
8. [Best Practices & Tips](#best-practices)

---

## Character Swapping with Seedance 2.5 {#character-swapping}

### What You'll Learn
Replace a person in an existing video with a specific portrait while preserving the original scene, framing, lighting, and audio.

### Prerequisites
- Source video (your B-roll or reference footage)
- Reference portrait image (the person you want to swap in)
- PayMeGPT with Seedance 2.5 access

### Step-by-Step Process

**1. Prepare Your Assets**
- Source video: Keep it short (15-30 seconds ideal for best quality)
- Reference portrait: Clean headshot, no hat/sunglasses, 512x512 or larger
- Ensure video and image have good lighting

**2. Upload to PayMeGPT Reference Library**
- Add both video and portrait image to your reference library
- Note the reference IDs returned

**3. Generate Character Swap**
```
Model: seedance-2-5
Input: referenceVideoUrls + referenceImageUrls
Prompt: "Replace the person in @video1 with the man in @image1. 
Keep the scene, camera, clothing, timing, and all audio identical."
Duration: Match source video length
```

**4. Result**
- New video with swapped character
- Original audio preserved
- Native soundtrack optional (can disable with generateAudio: false)

### Pro Tips
- Lighting consistency between portrait and video = better results
- Keep reference video 2-10 seconds for fastest processing
- Portrait should face forward for best swap quality
- Test with short clips before doing longer content

---

## Voice Cloning & Synthesis {#voice-cloning}

### Creating Your Cloned Voice

**Step 1: Clone Your Voice with ElevenLabs**
- Record 1-2 minutes of clear audio (you speaking naturally)
- Upload to ElevenLabs Voice Library
- Set name (e.g., "Chatbot King")
- Get your voice_id (e.g., `rrGArVIALAdsvGNidO9Y`)

**Step 2: Generate Speech from Text**
```
Model: eleven_english_sts_v2
Voice ID: your_cloned_voice_id
Input Text: Your script
Output: MP3 audio file
```

**Step 3: Apply to Video**
```
Source Video: Your character-swapped video
Audio: Generated speech MP3
Method: Merge with audioDucking=true
Background Volume: 0.35 (adjust to taste)
```

### Voice Quality Tips
- Speak clearly and naturally during recording
- Minimize background noise
- Use consistent tone and pace
- Test with short sentences first
- ElevenLabs quality improves with longer recordings

---

## Audio Production & Mixing {#audio-production}

### Layering Audio Elements

**Element 1: Primary Audio (Video Dialogue)**
- Keep your character's voice at 100% volume
- Ensure clear, audible speech

**Element 2: Background Music**
- Generate or upload instrumental track
- Keep tempo consistent with video pacing
- Use `musicDucking: true` to blend naturally
- Typical volume: 0.35-0.5 of primary audio

**Element 3: Sound Effects**
- Intro whoosh: 1-2 seconds, 1.0 volume
- Glitch effect: 0.5-2 seconds, 0.8 volume
- Place at strategic moments (intro, transitions, outro)

### Mixing Best Practices
- Primary dialogue should always be audible
- Music shouldn't overpower speech
- SFX add impact at key moments
- Test volume levels on phone speakers
- Leave 0.5s silence at start/end for breathing room

---

## Video Composition & Layout {#video-composition}

### Split-Screen Layouts

**Vertical Format (9:16 - Reels/TikTok)**
- Top half: Original video
- Bottom half: Character-swapped version
- Creates compelling before/after effect
- Perfect for social media

**Landscape Format (16:9 - YouTube/Web)**
- Left side: Original
- Right side: Swapped version
- Professional side-by-side comparison
- Better for presentations

### Generation Tips
- Use `aspectRatio: "9:16"` for vertical
- Use `aspectRatio: "16:9"` for landscape
- Merge method creates synchronized frames
- No black bars when using "fill" mode
- Keep both versions in sync perfectly

---

## Caption Generation & Sync {#caption-generation}

### Workflow

**1. Transcribe Video**
Use Whisper (speech-to-text) on your final video
- Get word-level timestamps
- Document timing precision

**2. Structure Captions**
- Break into 4-5 word phrases
- Match natural speech pauses
- Create timing segments

**3. Caption Specifications**
```json
{
  "style": "yellow (#FFFF00)",
  "background": "dark (semi-transparent black)",
  "alignment": "centered",
  "fontSize": 18,
  "maxWordsPerLine": 5
}
```

**4. Sync to Video**
```
Tool: merge_media or video editor
Input: Video + caption JSON with timestamps
Output: Final video with burned-in captions
```

### Caption Best Practices
- Keep short phrases (4-5 words max)
- Sync to word boundaries, not syllables
- Use bright colors (yellow/white) on dark backgrounds
- Test readability on mobile devices
- Add captions for accessibility

---

## Landing Page Lead Capture {#landing-page-setup}

### Page Structure

**Hero Section**
- Large, professional landscape image
- Clear headline: "Master AI Video Automation"
- Compelling subtitle describing value

**Video Embed**
- Embedded final video with playback controls
- Demonstrates your work in action
- Shows before/after or full workflow

**Lead Capture Form**
```
Fields:
- Full Name (required, text)
- Email Address (required, email)
- PayMeGPT Customer (select dropdown)
  Options: Yes / No, but interested / Thinking about it
```

**Success Flow**
1. User submits form
2. Data posts to webhook
3. Sheet auto-populates
4. Success message displays
5. Download button appears
6. Owner receives SMS alert

---

## Webhook Flows & Automation {#webhook-flows}

### Setting Up Your Capture Flow

**1. Create Flow in PayMeGPT**
```
Name: Character Swap Landing Page - Lead Capture
Enable Webhook: true
```

**2. Configure Actions**
- Append to Google Sheet (auto-sync)
- Send SMS to owner phone number
- Send confirmation to lead (optional)
- Record timestamp on each submission

**3. Sheet Mapping**
```
Columns:
- Timestamp → {{timestamp}} (auto-current time)
- Name → {{name}}
- Email → {{email}}
- PayMeGPT Customer → {{customer_status}}
- Status → "New"
```

**4. SMS Alert Setup**
```
Recipient: Your phone (E.164 format)
Message: "New lead: [Name] ([Email]). Customer: [Status]"
Timing: Immediately on submission
```

### Testing the Flow
1. Submit test form with your email
2. Check Google Sheet for new row
3. Verify you receive SMS alert
4. Confirm timestamp accuracy

---

## Best Practices & Tips {#best-practices}

### Content Creation
- **Lighting:** Consistent lighting between video and portrait = better swaps
- **Audio:** Clear speech with minimal background noise
- **Duration:** Keep videos 15-30 seconds for fast generation
- **Quality:** Higher resolution = better final output
- **Variety:** Create multiple videos to test different approaches

### Lead Capture Optimization
- **Form Fields:** Keep to 3-4 essential questions only
- **Mobile Friendly:** Test form on phones and tablets
- **Clear CTA:** Make download button obvious and actionable
- **Fast Loading:** Optimize images (compress PNG/JPG)
- **Trust Signals:** Show you've helped others (testimonials optional)

### Workflow Efficiency
- **Batch Process:** Create multiple videos before editing
- **Template Videos:** Use similar source clips for consistency
- **Voice Consistency:** Record multiple takes, use best for cloning
- **Library Organization:** Tag assets by project/client
- **Versioning:** Keep dated backups of working files

### Video Performance Tips
- **Captions Essential:** 85% of social video watched without sound
- **Hook Early:** First 3 seconds = most important
- **Pacing:** Match music tempo to dialogue pace
- **Call-to-Action:** Clear, specific ask (download, visit, subscribe)
- **Mobile First:** Design for 9:16 vertical format primarily

### Troubleshooting

**Character Swap Not Good?**
- Try different lighting on portrait
- Use clearer, higher-contrast headshot
- Ensure video shows full body/shoulders
- Consider longer source video (2-5 seconds min)

**Audio Out of Sync?**
- Regenerate with same duration
- Check video frame rate (30fps ideal)
- Reduce video length by 0.5-1 second
- Try different audio merge settings

**Lead Flow Not Triggering?**
- Verify webhook URL in landing page form
- Check flow is active (enabled toggle)
- Test webhook directly with curl/Postman
- Confirm sheet has proper headers

**SMS Not Arriving?**
- Verify phone number in E.164 format (+15551234567)
- Check Twilio credits/setup
- Confirm flow action has SMS enabled
- Test with manual SMS action first

---

## Advanced Techniques

### Multi-Person Videos
- Use multiple reference images
- Swap different people in different scenes
- Layer multiple character swap outputs

### Long-Form Content
- Break into 30-second segments
- Swap each segment independently
- Stitch together in final edit

### Interactive Landing Pages
- Add multiple CTA buttons (download, demo, book call)
- Use form data to personalize next steps
- Track conversion rates by source

### Scaling Operations
- Create flow templates for each campaign
- Batch form submissions for efficiency
- Use Google Sheet data for follow-up sequences
- Automate post-conversion workflows

---

## Resources & Tools

**Primary Platforms**
- PayMeGPT: Video generation, lead capture, automation
- Seedance 2.5: Character swap (via PayMeGPT)
- ElevenLabs: Voice cloning & synthesis
- Google Sheets: Lead data storage & CRM

**Testing Tools**
- Twilio Console: SMS testing
- Postman: Webhook testing
- Browser DevTools: Form debugging
- Mobile Device: Real-world testing

**Learning Resources**
- PayMeGPT Documentation
- Seedance Technical Guides
- ElevenLabs Voice Cloning Tutorial
- Video Marketing Best Practices

---

## Next Steps

1. **Record & Clone Your Voice**
   - Record 1-2 minutes of natural speech
   - Clone in ElevenLabs
   - Save voice_id for future use

2. **Create Your First Swap**
   - Find or record short source video
   - Take clean portrait photo
   - Generate character swap in PayMeGPT
   - Review and iterate

3. **Build Your Landing Page**
   - Create hero image
   - Write compelling copy
   - Set up lead capture form
   - Configure webhook flow

4. **Launch & Iterate**
   - Publish landing page
   - Track submissions in Google Sheet
   - Analyze performance metrics
   - Refine based on feedback

5. **Scale & Automate**
   - Create multiple campaigns
   - Build email sequences for leads
   - Set up follow-up automations
   - Track ROI and optimize

---

## Conclusion

You now have the complete blueprint for creating professional AI-powered video content and capturing leads at scale. The combination of character swaps, voice cloning, and automated lead capture creates a powerful content and sales funnel.

Start with one video, validate your process, then scale to multiple campaigns. Each iteration will improve your results and efficiency.

**Happy creating! 🚀**

---

*Last Updated: August 20, 2026*
*Version: 1.0*
*Created by: Chatbot King (PayMeGPT)*