Spotter

AI-assisted sound design spotting for macOS. Analyze any video → get a complete cue sheet with timecodes → generate or search SFX → export DAW-synchronized stems.

Built by Lemö Labs.

---

What it is

Two things in one repo:

• Electron app — visual timeline for manual spotting, clip editing, gain, fades, and export
• MCP server — automated pipeline for use directly from Claude Code (no GUI needed)

Both share the same project format (analysis.json + media/ + export/).

---

Electron app (macOS)

Requirements

• macOS (Apple Silicon)
• Node.js 18+

Install & run

git clone https://github.com/GuilleMcDugan/spotter.git
cd spotter
npm install
npm run dev

Build DMG

npm run dist
# → dist/SPOTTER-1.0.0-arm64.dmg

Features

• AI video analysis (extracts frames → Claude → cue sheet)
• Multi-track timeline with drag, trim, vertical clip movement, split (S key)
• Per-clip gain and fade in/out (drag handles)
• ElevenLabs SFX generation + Freesound search, built in
• Export: WAV 16-bit / 24-bit / MP3 320k at 44.1kHz or 48kHz
• DAW-synchronized stems (each file padded to full video length)
• Auto-save timeline between sessions
• J/K/L transport + frame-accurate scrubbing

---

MCP server (Claude Code integration)

See mcp-server/README.md for setup and tool reference.

Quick start:

cd mcp-server
npm install && npm run build

Add to ~/.claude/settings.json:

{
  "mcpServers": {
    "spotter": {
      "command": "node",
      "args": ["/absolute/path/to/spotter/mcp-server/dist/index.js"]
    }
  }
}

---

Guía de instalación

Documentación completa en Notion (español): SPOTTER — Guía de instalación

---

API keys

Keys are stored locally — never in the repo.

• Electron app: Settings panel (⚙) → paste keys → stored in system keychain via electron-store
• MCP server: configure_provider({ provider: "elevenlabs", api_key: "..." }) → stored in ~/.spotter/config.json

Required:

• ElevenLabs — SFX generation
• Freesound — free SFX search

Claude API key: handled by the Claude CLI (claude auth), not by SPOTTER.

---

License

MIT