Configuration

All settings are controlled via environment variables or ~/.openclaw/openclaw.json.

Environment Variables

Variable	Default	Description
VIDEO_VISION_API_KEY	required	Vision model API key
VIDEO_VISION_API_URL	https://api.openai.com/v1/chat/completions	Any OpenAI-compatible vision endpoint
VIDEO_VISION_MODEL	gpt-4o	Vision model to use
VIDEO_VISION_MODE	auto	Extraction mode: auto / ytdlp / browser (see Extraction Modes)
VIDEO_VISION_PROXY	—	Default proxy URL (HTTP/HTTPS/SOCKS5)
VIDEO_VISION_FRAME_INTERVAL	5	Seconds between extracted frames
VIDEO_VISION_MAX_FRAMES	20	Maximum frames per video
VIDEO_VISION_COOKIES_DIR	—	Directory containing cookie files (see Cookies)
VIDEO_VISION_LOW_RESOURCE	false	Skip resource checks, disable transcription
VIDEO_VISION_TRANSCRIPTION	auto	Transcription mode: auto / on / off (auto = on unless low-resource)
VIDEO_VISION_WHISPER_PATH	whisper-cli	Path to whisper-cli binary
VIDEO_VISION_WHISPER_MODEL_PATH	(auto-detect)	Full path to ggml model file
VIDEO_VISION_WHISPER_MODEL	medium	Model name: tiny/base/small/medium/large-v3
VIDEO_VISION_WHISPER_THREADS	0 (auto)	CPU threads for whisper (0 = cores/2)
VIDEO_VISION_WHISPER_LANGUAGE	auto	Audio language hint
VIDEO_VISION_BROWSER	local	Browser backend: local / browserless / browserbase / steel
VIDEO_VISION_BROWSERLESS_TOKEN	—	Browserless API token
VIDEO_VISION_BROWSERBASE_API_KEY	—	Browserbase API key
VIDEO_VISION_BROWSERBASE_PROJECT_ID	—	Browserbase project ID
VIDEO_VISION_STEEL_API_KEY	—	Steel API key

OpenClaw JSON Config

{
  "skills": {
    "entries": [
      {
        "name": "video-vision",
        "env": {
          "VIDEO_VISION_API_KEY": "sk-...",
          "VIDEO_VISION_MODEL": "gpt-4o",
          "VIDEO_VISION_MODE": "ytdlp",
          "VIDEO_VISION_PROXY": "http://127.0.0.1:7890",
          "VIDEO_VISION_FRAME_INTERVAL": "5",
          "VIDEO_VISION_MAX_FRAMES": "20",
          "VIDEO_VISION_COOKIES_DIR": "~/.openclaw/cookies"
        }
      }
    ]
  }
}

Proxy

Supports HTTP, HTTPS, and SOCKS5 proxies. The proxy is used for:

• yt-dlp video metadata & download
• Browser network traffic (Phase 2)

export VIDEO_VISION_PROXY="http://127.0.0.1:7890"
# or
export VIDEO_VISION_PROXY="socks5://127.0.0.1:1080"

Per-request proxy via CLI flag:

node src/index.js https://youtube.com/watch?v=xxx --proxy=http://127.0.0.1:7890

Vision API Endpoints

Any OpenAI-compatible /v1/chat/completions endpoint works:

# OpenAI (default)
export VIDEO_VISION_API_URL="https://api.openai.com/v1/chat/completions"
export VIDEO_VISION_MODEL="gpt-4o"

# Anthropic (via compatible proxy)
export VIDEO_VISION_API_URL="https://your-proxy/v1/chat/completions"
export VIDEO_VISION_MODEL="claude-sonnet-4-20250514"

# Local model
export VIDEO_VISION_API_URL="http://localhost:11434/v1/chat/completions"
export VIDEO_VISION_MODEL="llava"