ai-video-production-master

这段内容是审核时保存的快照。外部来源才是完整且最新的版本。

---
name: ai-video-production-master
description: Expert in script-to-video production pipelines for Apple Silicon Macs. Specializes in hybrid local/cloud workflows, LoRA training for character consistency, motion graphics generation, and
  artist commissioning. Activate on 'AI video production', 'script to video', 'video generation pipeline', 'character consistency', 'LoRA training', 'cloud GPU', 'motion graphics', 'Wan I2V', 'InVideo alternative'.
  NOT for real-time video editing, video compositing (use DaVinci/Premiere), audio production, or 3D modeling (use Blender/Maya).
allowed-tools: Read,Write,Edit,Bash(python:*,ffmpeg:*,npm:*),WebFetch,mcp__firecrawl__firecrawl_search
metadata:
  category: AI & Machine Learning
  pairs-with:
  - skill: sound-engineer
    reason: Audio for AI-generated videos
  - skill: voice-audio-engineer
    reason: Voice synthesis for narration
  tags:
  - video
  - ai-generation
  - lora
  - cloud-gpu
  - motion-graphics
  - comfyui
---

# AI Video Production Master

Expert in script-to-video production pipelines for Apple Silicon Macs. Specializes in:
- **Multiple video approaches**: Stock footage, T2V (Sora-style), I2V, hybrid
- Hybrid local/cloud workflows for cost optimization
- Style and character consistency (LoRA, IPAdapter, prompt discipline)
- Motion graphics and synthetic elements (title cards, data viz, lower thirds)
- Artist commissioning for training datasets
- Cloud GPU orchestration (Vast.ai, RunPod)

## When to Use

✅ **USE this skill for:**
- Script-to-video production pipelines
- Stock footage assembly (InVideo-style workflows)
- Text-to-video generation (Sora, Runway, Pika, Kling)
- Image-to-video animation (Wan I2V, ComfyUI)
- Cloud GPU orchestration (Vast.ai, RunPod, Lambda)
- Motion graphics generation (title cards, lower thirds, data viz)
- LoRA training for character/style consistency
- Artist commissioning for training datasets
- Cost optimization between local and cloud processing

❌ **DO NOT use for:**
- Real-time video editing → use DaVinci Resolve, Premiere Pro
- Video effects/compositing → use After Effects, Fusion
- Audio production/mixing → use `sound-engineer` skill
- 3D modeling/animation → use Blender, Maya, or `physics-rendering-expert` skill
- Static image generation → use `clip-aware-embeddings` or image gen tools

## Video Generation Approaches

Choose the right approach based on your content:

### Stock Footage (Invideo-style) - RECOMMENDED for most content
Best for: Educational, corporate, explainers, documentaries
- Uses curated stock libraries (Pexels, Pixabay, Storyblocks)
- Most professional, reliable results
- Fast turnaround (~30 min for full video)
- Script → AI selects matching clips → voiceover + music
```bash
python scripts/stock_video_generator.py --script script.txt --style documentary
```

### Text-to-Video (Sora-style) - For creative/artistic content
Best for: Abstract visuals, creative shorts, unique scenes
- True generative AI (no stock footage)
- Uses: Sora API, Runway Gen-3, Pika, Kling
- Cleaner than I2V (no weird image artifacts)
- Storyboard control for multi-shot narratives
```bash
python scripts/t2v_generator.py --prompt "A serene mountain lake at sunset" --provider sora
```

### Image-to-Video (I2V) - For animating specific images
Best for: Animating logos, concept art, specific compositions
- Animates existing images with subtle motion
- Can look "weird" if source images are AI-generated
- Best with clean, professional source images
```bash
python scripts/cloud_i2v_batch.py --images ./keyframes --provider vastai
```

### Hybrid Approach
Combine approaches per shot:
- Shot 1-3: Stock footage (b-roll, establishing)
- Shot 4-5: T2V (creative transitions)
- Shot 6-10: Stock footage (talking head, outro)

## Key Capabilities

### 1. Cost Optimization
Compare and recommend the optimal mix of local (M4 Max) vs cloud (H100/A100) processing:
```bash
python scripts/cost_calculator.py --shots 10 --duration 5
```

### 2. Cloud Batch Processing