veo-3.1-tutorial

KO | EN

gitlite — search

#ai #typescript #llm #go #machine-learning #rust #deep-learning #macos #java #cli #kubernetes #vue #pytorch #ios

veo-3.1-tutorial

(★ 1)

Automated video generation with Google's VEO 3.1 using agentic dropzone patterns and Claude Code SDK

veo-3.1-tutorial 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# VEO 3.1 Agentic Dropzone Tutorial

Automated video generation with Google's VEO 3.1 using agentic dropzone patterns and Claude Code SDK.

This tutorial demonstrates how to create AI-powered video workflows that automatically trigger when you drop text files into monitored directories.

## Features

- **Character Prompt Factory**: Generate character-consistent video prompts with AI
- **Basic Video Generation**: Text-to-video with VEO 3.1
- **Reference Images**: Guide video generation with up to 3 reference images
- **Automated Workflows**: File watcher triggers Claude Code SDK to execute generation scripts
- **Multi-zone Support**: Separate directories for different workflow types
- **Utility Functions**: Image resizing, video downloading, and prompt templates

## Prerequisites

### 1. Google AI API Key

Get your API key from [Google AI Studio](https://aistudio.google.com/apikey)

### 2. Anthropic API Key (for Character Factory)

Get your API key from [Anthropic Console](https://console.anthropic.com/)

Required for the Character Prompt Factory to generate character-consistent video prompts.

### 3. Claude Code CLI

Install the Claude Code CLI globally:

```bash
npm install -g @anthropic-ai/claude-code
```

Verify installation:

```bash
claude --version
```

### 4. Python Environment

- Python 3.8 or higher
- pip package manager

### 5. Node.js

- Node.js 18 or higher (for Claude Code CLI)
- npm or nvm

## Installation

### 1. Clone the Repository

```bash
git clone <your-repo-url>
cd veo-3.1-tutorial
```

### 2. Install Python Dependencies

```bash
pip install -r requirements.txt
```

### 3. Configure Environment Variables

Copy the example environment file:

```bash
cp .env.example .env
```

Edit `.env` and add your Google API key:

```bash
GOOGLE_API_KEY=your_google_api_key_here
```

### 4. Verify Claude Code Setup

Make sure Claude Code is accessible in your PATH. If using nvm:

```bash
# Find your nvm bin directory
which claude

# Should return something like:
# /Users/yourname/.nvm/versions/node/v23.9.0/bin/claude
```

## Usage

### Starting the Drop Zone Watcher

Run the watcher to monitor all configured drop zones:

```bash
python drop_zone_watcher.py
```

The watcher will:
- Create zone directories if they don't exist
- Monitor for new `.txt` files
- Trigger Claude Code SDK workflows automatically
- Display colored status updates for each zone

### Character Prompt Factory (Optional)

The Character Prompt Factory helps you generate detailed, character-consistent video prompts using Claude AI.

**Quick Start:**

```bash
cd character_factory
python character_prompt_agent.py nyx_shadow_hacker
```

**Interactive Session:**

```
🎬 Character Prompt Factory - VEO 3.1 Video Generator
Character: Nyx the Shadow Hacker
Style: Blade Runner meets Ghost in the Shell; neon-lit cyberpunk

Commands: /quit (exit) | /character (show details)
        /new (reset context) | /extend (toggle extension mode)

[Turn 1] Nyx the Shadow Hacker > Nyx hacks into a corporate mainframe
🤔 Generating VEO 3.1 prompt...
```

**Available Commands:**
- `/quit` - Exit the agent
- `/character` - Show current character details
- `/new` - Start fresh conversation (clear context)
- `/extend` - Toggle extension mode for video continuation

**Workflow:**
1. Run the character factory with a character name
2. Chat with the AI to generate prompts
3. Copy the generated prompt
4. Save to a `.txt` file in a dropzone directory
5. Watch the drop zone watcher generate your video!

See [`character_factory/README.md`](character_factory/README.md) for more details including adding custom characters.

### Drop Zone 1: Basic Video Generation

**Zone Directory**: `veo3_direct_zone/`

Create a text file with your video prompt:

```bash
echo "A serene beach at sunset with waves gently lapping the shore" > veo3_direct_zone/beach.txt
```

The system will:
1. Detect the new file
2. Trigger the VEO 3.1 generation script
3. Generate a 9:16 portrait video (720p, 8 seconds)
4. Save output to `veo3_direct_zone/outputs/`

**Video Extension (Optional)**:

The basic dropzone script also supports extending videos. To use this feature, add extension prompts to your text file:

```
prompt: A serene beach at sunset with waves gently lapping the shore
extend1: The camera pans right to reveal a hidden waterfall
extend2: Wildlife appears near the water's edge
```

Each `extend1:`, `extend2:`, etc. will create an additional 8-second video segment continuing from the previous one.

### Drop Zone 2: Reference Image Video Generation

**Zone Directory**: `veo3_reference_zone/`

Create a prompt file with reference images:

```
prompt: A mystical forest transforming through the seasons
reference_images:
- path/to/spring_forest.jpg
- path/to/autumn_forest.jpg
```

Save to `veo3_reference_zone/forest.txt`

The system will:
1. Detect the file and parse references
2. Upload images to Google AI
3. Generate a 16:9 video guided by the images (720p, 8 seconds)
4. Save output to `veo3_reference_zone/outputs/`

**Important**: Reference images require:
- Aspect ratio: 16:9
- Duration: 8 seconds
- Up to 3 images maximum

## Configuration

### drops.yaml

The `drops.yaml` file configures all drop zones:

```yaml
drop_zones:
  - name: "Veo 3.1"
    zone_dirs:
      - "veo3_direct_zone"
    file_patterns:
      - "*.txt"
    reusable_prompt: ".claude/commands/veo3_direct.md"
    events:
      - "created"
    agent: "claude_code"
    model: "sonnet"
    color: "bright_green"
    create_zone_dir_if_not_exists: true
```

**Key Parameters**:
- `zone_dirs`: Directories to monitor
- `file_patterns`: File types to watch for
- `reusable_prompt`: Claude prompt file that orchestrates the workflow
- `events`: File system events to trigger on (`created`, `modified`, `deleted`)
- `color`: Terminal color for status messages

### Claude Command Files

Located in `.claude/commands/`, these files contain the prompts that instruct Claude Code SDK on how to execute the generation scripts:

- `veo3_direct.md` - Basic video generation workflow
- `veo3_reference.md` - Reference image workflow

## Project Structure

```
veo-3.1-tutorial/
├── .claude/
│   └── commands/              # Claude Code SDK prompt files
│       ├── veo3_direct.md
│       └── veo3_reference.md
├── character_factory/         # Character-consistent prompt generation
│   ├── character_prompt_agent.py  # Main agent script
│   ├── character_config.yaml      # Character definitions
│   ├── system_prompts.yaml        # Claude instruction templates
│   ├── Transparent_Logo.png       # Logo asset
│   ├── .env.example               # Environment template
│   └── README.md                  # Character factory docs
├── scripts/                   # Generation scripts
│   ├── veo3_dropzone.py      # Basic video generation
│   └── veo3_reference_dropzone.py  # Reference image generation
├── utils/                     # Utility functions
│   ├── __init__.py           # Package initialization
│   ├── director.py           # Director utilities
│   ├── downloader.py         # Video download helpers
│   ├── prompt_templates.py   # Prompt style templates
│   └── resizer.py            # Image resizing utilities
├── examples/                  # Example prompt files
├── veo3_direct_zone/         # Basic video drop zone (auto-created)
│   └── outputs/
├── veo3_reference_zone/      # Reference image drop zone (auto-created)
│   └── outputs/
├── drop_zone_watcher.py      # Main watcher process
├── drops.yaml                # Drop zone configuration
├── requirements.txt          # Python dependencies
├── .env                      # Environment variables (create from .env.example)
├── .env.example              # Environment template
├── .gitignore
└── README.md
```

## Technical Details

### VEO 3.1 Model

- **Model ID**: `veo-3.1-generate-preview`
- **Supported Aspect Ratios**: 9:16, 16:9
- **Resolutions**: 720p, 1080p
- **Duration**: 8 seconds per generation
- **Reference Images**: Up to 3 images, asset type

### Claude Code SDK Integration

The watcher uses Claude Code SDK's Python bindings to:
1. Parse prompt files
2. Execute generation scripts with proper error handling
3. Manage API credentials and rate limiting
4. Provide rich terminal output

### File Watcher

Built with `watchdog` library for cross-platform file system monitoring:
- Monitors multiple directories simultaneously
- Debounces rapid file changes
- Handles file creation, modification, deletion events
- Colorized terminal output with `rich` library

## Troubleshooting

### Claude Code Not Found

If you get `Claude Code not found` error:

1. Verify installation:
```bash
npm list -g @anthropic-ai/claude-code
```

2. Check PATH:
```bash
echo $PATH | grep nvm
```

3. Reinstall if needed:
```bash
npm install -g @anthropic-ai/claude-code
```

### Google API Errors

If you get authentication errors:

1. Verify API key in `.env` file
2. Check API key is valid at [Google AI Studio](https://aistudio.google.com/apikey)
3. Ensure API key has VEO 3.1 access enabled

### Reference Image Errors

Common issues:
- **Too many images**: Maximum 3 reference images
- **Wrong aspect ratio**: Must use 16:9 for reference images
- **Wrong duration**: Must use 8 seconds for reference images

Fix by ensuring your prompt uses:
```python
ASPECT_RATIO = "16:9"
DURATION_SECONDS = 8
```

## Examples

See the `examples/` directory for sample prompt files demonstrating:
- Basic text-to-video prompts
- Reference image formatting
- Advanced prompting techniques

## License

MIT

## Credits

Built with:
- [Google VEO 3.1 API](https://ai.google.dev/gemini-api/docs/video)
- [Claude Code SDK](https://docs.anthropic.com/claude/docs)
- [Watchdog](https://github.com/gorakhargosh/watchdog)
- [Rich](https://github.com/Textualize/rich)

veo-3.1-tutorial

veo-3.1-tutorial 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

veo-3.1-tutorial

veo-3.1-tutorial 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README