Video Asset Manager

Video Asset Manager

Source, download, and organize video clips for music video production using Pexels API and GenSpark AI generation.

Workflow

1. Input: Scene Plan JSON

Receive scene plan from creative-director skill:

{
  "scenes": [
    {
      "sceneNumber": 1,
      "clipKeywords": ["sunrise", "mountains", "golden hour"],
      "duration": 15
    }
  ]
}

2. Search Pexels for Stock Footage

Pexels API Access:

• Credentials: Ask McKinzie for Pexels API key
• Endpoint: https://api.pexels.com/videos/search
• Rate limit: 200 requests/hour (free tier)

Search strategy:

1. For each scene, search using clipKeywords
2. Filter by:
   • Orientation (landscape for YouTube, square/vertical for social)
   • Duration (prefer clips ≥ scene duration)
   • Quality (HD minimum, 4K preferred)
3. Download top 2-3 options per scene for flexibility

Pexels search script: See scripts/search_pexels.py

3. Generate AI Clips with GenSpark (PREFERRED METHOD)

When to use GenSpark:

• Abstract concepts (faith, hope, spiritual themes)
• Scenes where stock footage doesn't match
• Unique/custom visuals not available in stock libraries
• Cinematic quality needed (better than stock footage)
• Full music video production (all clips AI-generated)

⚡ BATCH WORKFLOW (FAST - USE THIS!):

See full documentation: references/genspark-batch-workflow.md

1. Prepare all prompts in numbered list format:

   Generate [N] videos using PixVerse model, 16:9 aspect ratio, 4K resolution:
   
   1. [Detailed cinematic prompt for scene 1]
   2. [Detailed cinematic prompt for scene 2]
   ... etc
   

2. Submit to GenSpark:

   • Go to https://www.genspark.ai (main homepage)
   • Paste ENTIRE prompt list into main textbox
   • Press Enter

3. Confirm generation:

   • GenSpark will ask to confirm
   • Reply: "A) Proceed with all [N] videos now. Use 8 second duration, 16:9 aspect ratio."

4. Wait for completion:

   • PixVerse: ~30 seconds per video
   • 21 videos ≈ 15-20 minutes total
   • Videos generate in parallel

5. Download all clips when complete

Model: PixVerse (official/pixverse/v5) - fast, high quality Duration: 5s or 8s (prefer 8s for music videos) Aspect: 16:9 landscape, 9:16 for shorts

Prompt tips for cinematic quality:

• Include camera movement: "slow push-in", "overhead drift"
• Specify lighting: "golden hour", "soft window backlight"
• Add film aesthetic: "35mm film grain", "shallow depth of field"
• Describe mood: "contemplative", "nostalgic", "ethereal"

4. Organize Assets

Folder structure:

project-name/
├── assets/
│   ├── scene-01-sunrise-mountains.mp4
│   ├── scene-02-walking-forest.mp4
│   ├── scene-03-ai-faith-abstract.mp4
│   └── scene-04-ocean-waves.mp4
├── scene-plan.json
└── asset-manifest.json

Asset manifest:

{
  "projectName": "Song Title - Music Video",
  "createdAt": "2026-01-29",
  "assets": [
    {
      "sceneNumber": 1,
      "filename": "scene-01-sunrise-mountains.mp4",
      "source": "pexels",
      "pexelsId": "12345678",
      "duration": 18,
      "resolution": "1920x1080",
      "keywords": ["sunrise", "mountains", "golden hour"]
    },
    {
      "sceneNumber": 3,
      "filename": "scene-03-ai-faith-abstract.mp4",
      "source": "genspark",
      "prompt": "abstract light rays representing faith and hope",
      "duration": 10,
      "resolution": "1920x1080"
    }
  ]
}

5. Quality Check

Before passing to music-video-producer:

• All scenes have at least 1 clip sourced
• Clip durations ≥ scene duration (allow trimming)
• Resolution is consistent (1080p minimum)
• Files downloaded and playable
• Asset manifest generated
• Folder organized and labeled

Scripts

scripts/search_pexels.py

Search and download clips from Pexels:

python scripts/search_pexels.py \
  --scene-plan scene-plan.json \
  --output-dir assets/ \
  --api-key $PEXELS_API_KEY \
  --top-n 2

See scripts/search_pexels.py for implementation.

scripts/generate_genspark.py

Generate AI clips with GenSpark:

python scripts/generate_genspark.py \
  --scene-plan scene-plan.json \
  --scenes 3,5,7 \
  --output-dir assets/

See scripts/generate_genspark.py for implementation.

Reference Files

• references/pexels-api-docs.md - Pexels API usage and examples
• references/genspark-prompts.md - Effective prompts for AI video generation
• references/genspark-batch-workflow.md - Batch video generation process
• references/genspark-download-workflow.md - Downloading from AI Drive
• references/video-quality-standards.md - Resolution, format, codec requirements

Deliverables

1. Downloaded video clips - Organized in assets/ folder
2. Asset manifest JSON - Metadata for each clip
3. Coverage report - Which scenes have clips, which need alternatives

Error Handling

If Pexels search returns no results:

• Try alternate keywords (synonyms)
• Expand search to related concepts
• Flag scene for GenSpark AI generation

If clip quality is low:

• Search for HD/4K alternatives
• Consider upscaling with AI (Topaz Video AI)
• Flag for manual review

If API rate limit hit:

• Queue remaining searches
• Resume after cooldown period
• Report estimated completion time