TikTok AI Model Video Generator

Create AI-generated fashion models wearing your products and animate them into engaging TikTok livestream videos. This workflow combines multiple AI tools to produce realistic, professional product showcase content in under 5 minutes.

Quick Start

Complete 4-step workflow to generate AI model video:

# Step 1: Select Pinterest reference
# Find a fashion pose/angle you like on Pinterest

# Step 2: Generate JSON prompt with Claude
# Ask Claude: "Give me detailed JSON prompt for this image holding [your product]"

# Step 3: Generate image with Nano Banana Pro
# Paste Claude's JSON prompt with your product image (white background)

# Step 4: Animate video with Veo or Kling
# Upload generated image to Veo/Kling and animate

# Total time: Under 5 minutes

Workflow Steps

Step 1: Pinterest Reference Selection

Goal: Find high-quality fashion poses that match your product's aesthetic.

Action: Browse Pinterest and save 2-3 reference images showing:

• Desired pose (standing, walking, sitting, etc.)
• Camera angle (full body, close-up, profile, etc.)
• Lighting style (studio, natural, dramatic)
• Background preference (clean, lifestyle, minimalist)

Tips:

• Search: "fashion model pose", "product photography pose", "[your product] model"
• Look for consistent brand aesthetic
• Save URLs or download images for Claude reference

---

Step 2: Claude JSON Prompt Generation

Goal: Create detailed, structured prompt for Nano Banana Pro.

Prompt Template:

Give me detailed JSON prompt for this Pinterest image holding [PRODUCT DESCRIPTION]:

{
  "subject": {
    "description": "[Detailed product description]",
    "pose": "[Pose from Pinterest]",
    "angle": "[Camera angle]",
    "lighting": "[Lighting style]"
  },
  "model": {
    "appearance": "[Model physical description]",
    "outfit": "[Clothing/style details]",
    "expression": "[Facial expression]"
  },
  "environment": {
    "background": "[Background description]",
    "location": "[Setting/context]",
    "atmosphere": "[Mood/vibe]"
  },
  "technical": {
    "style": "[Photography style]",
    "camera": "[Camera settings]",
    "resolution": "[Image resolution]"
  }
}

Use the Pinterest image as visual reference for pose and composition.

Claude Capabilities:

• Analyze Pinterest image composition
• Extract pose, angle, lighting details
• Integrate product naturally into scene
• Generate JSON formatted for Nano Banana Pro

Product Image Requirements:

• White or neutral background
• High resolution (minimum 1024x1024)
• Clear product visibility
• Professional lighting
• No text or watermarks

---

Step 3: Nano Banana Pro Image Generation

Goal: Generate photorealistic AI model wearing your product.

Prerequisites:

• Nano Banana Pro access (https://higgsfield.ai)
• Higgsfield account for tool access
• API key or web interface access

Process:

1. Open Nano Banana Pro (Higgsfield)
2. Upload your product image (white background)
3. Paste Claude's JSON prompt
4. Adjust parameters:
   • Resolution: 1024x1024 (standard)
   • Style: Photorealistic
   • Quality: High
   • Variation: Generate 2-3 options
5. Click Generate
6. Select best result

Expected Output:

• AI model in pose matching Pinterest reference
• Product naturally integrated into scene
• Photorealistic quality (professional photography)
• Consistent lighting and shadows
• White/neutral background (easier for video editing)

Troubleshooting:

• Product not visible: Check JSON prompt includes product description clearly
• Pose mismatch: Add more specific pose instructions to JSON
• Unrealistic: Lower "style strength" or adjust "model appearance" details
• Lighting issues: Specify "studio lighting" in environment section

---

Step 4: Veo/Kling Video Animation

Goal: Animate generated image into engaging TikTok video.

Tool Options:

Veo (Google's AI Video Generator)

• Access: Via Higgsfield platform
• Input: Generated image from Nano Banana Pro
• Output: Animated video (3-10 seconds)
• Features:
  • Natural movement
  • Product-focused animation
  • High quality (1080p+)

Animation Prompts:

Prompt ideas for Veo:
- "Subtle body sway, arms gently moving, natural breathing motion"
- "Model turning slightly to show product from different angles"
- "Small hand gestures highlighting product features"
- "Natural head movement, facial expression changes"
- "Walking slowly, product clearly visible"

Kling AI Video Generator

• Alternative to Veo
• Similar workflow
• May offer different animation styles
• Check which produces better results for your use case

Process:

1. Upload Nano Banana Pro generated image
2. Choose animation style (subtle, dynamic, product-focused)
3. Enter animation prompt (see above)
4. Generate 3-5 second video
5. Review and refine if needed
6. Export for TikTok upload

Video Settings:

• Duration: 3-5 seconds (optimal for TikTok)
• Resolution: 1080x1920 (9:16 vertical)
• Frame rate: 24-30 fps
• Format: MP4 (TikTok compatible)

---

Use Cases

E-commerce Product Showcase

• Perfect for: Fashion, jewelry, accessories, cosmetics
• Workflow: Generate AI models in multiple poses
• Output: Product-focused videos showcasing features
• Time: 5 minutes per video (vs. hours of traditional shoots)

TikTok Livestream Content

• Use case: 24/7 AI model livestreams (as mentioned by @barkmeta)
• Workflow: Generate variations, animate, loop
• Advantage: Scalable, no human models needed
• Platforms: TikTok, Instagram, YouTube Shorts

Social Media Marketing

• Platforms: TikTok, Instagram Reels, YouTube Shorts
• Content types:
  • Product launches
  • Feature highlights
  • Seasonal collections
  • A/B testing different styles

---

Optimization Tips

Better Results

1. Pinterest Reference Quality:

   • Choose high-resolution images
   • Match your brand aesthetic
   • Consider lighting conditions

2. Prompt Specificity:

   • Be detailed in JSON structure
   • Include lighting, camera, and style
   • Reference Pinterest image explicitly

3. Product Preparation:

   • Clean, white background essential
   • Professional photography quality
   • Multiple angles available

4. Animation Subtlety:

   • Natural movements (no jerky motions)
   • Focus on product visibility
   • Keep videos short (3-5 seconds)

Time Optimization

Batch Workflow:

1. Generate 10-20 images at once (Nano Banana Pro batch)
2. Animate top 3-5 selections
3. Create content calendar (week/month)
4. Schedule TikTok posts

Tools Access:

• Save Claude JSON prompts as templates
• Reuse successful prompts for similar products
• Build prompt library for faster iteration

---

Common Issues & Solutions

---

Advanced Techniques

Multi-Pose Product Showcase

1. Generate 3 different poses (front, side, detail)
2. Animate each individually
3. Combine into longer TikTok video
4. Add transitions between poses

A/B Testing

1. Create 2-3 variations of same product
2. Test different poses/backgrounds
3. Compare engagement metrics
4. Optimize for best performing style

Seasonal Collections

1. Update JSON prompts for seasonal context
2. Change lighting (warm for summer, cool for winter)
3. Adjust background themes
4. Generate themed video sets

---

Required Tools

All tools accessible via Higgsfield:

1. Claude AI - JSON prompt generation
2. Nano Banana Pro - Photorealistic image generation
3. Veo 3.1 - Video animation
4. Kling AI - Alternative video generator

Alternative Workflows:

• Replace Claude with GPT-4 (if available)
• Replace Nano Banana with Midjourney/DALL-E
• Replace Veo with Runway ML/Pika Labs

---

Cost & Time Estimate

Per Video:

• Time: 3-5 minutes
• Tools: Free tier available (check limits)
• Commercial use: Verify tool licensing terms

Batch Production:

• 10 videos: ~30-60 minutes
• 50 videos: ~2.5-4 hours
• Scale: Unlimited (with tool API access)

Comparison:

• Traditional photoshoot: $500-$2,000+ per day
• AI workflow: Free to $50/month (subscription)
• Time savings: 95%+ reduction

---

Troubleshooting

Tool Access Issues

• Higgsfield account: Create free account at https://higgsfield.ai
• API rate limits: Check free tier limits, consider upgrade
• Login problems: Clear browser cache, try different browser

Quality Issues

• Low resolution: Increase image resolution to 2048x2048
• Artifacts: Regenerate with different random seed
• Inconsistent style: Use same JSON prompt template for batch

Animation Problems

• Jerky motion: Simplify animation prompt, use "subtle" keywords
• Product out of frame: Add "keep product in frame" to prompt
• Too fast: Reduce movement in prompt, extend duration

---

Examples

Jewelry Product

Pinterest: Minimalist gold necklace, model looking down
Product: Gold chain necklace on white background
Claude Prompt: "Generate JSON for pendant necklace, model looking down, studio lighting"
Nano Banana: Photorealistic close-up
Veo Animation: "Gentle sway, necklace catching light"

Fashion Item

Pinterest: Full body fashion pose, walking stance
Product: White t-shirt, lifestyle setting
Claude Prompt: "Generate JSON for casual wear, walking motion, outdoor lighting"
Nano Banana: Full body photorealistic
Veo Animation: "Natural walking motion, arms gently swinging"

Accessory

Pinterest: Hand holding phone case, focus on product
Product: Designer phone case
Claude Prompt: "Generate JSON for phone case in hand, close-up, clean background"
Nano Banana: High detail macro shot
Veo Animation: "Subtle hand movement, showing product angles"

---

File Structure

tiktok-ai-model-generator/
├── SKILL.md (this file)
├── scripts/
│   └── generate_workflow.py (optional automation script)
└── references/
    ├── pinterest_tips.md (Pinterest selection guide)
    └── prompt_templates.md (Reusable JSON templates)

See bundled scripts and references for automation tools and detailed templates.