Speak

                          ███████╗██████╗ ███████╗ █████╗ ██╗  ██╗
                          ██╔════╝██╔══██╗██╔════╝██╔══██╗██║ ██╔╝
                          ███████╗██████╔╝█████╗  ███████║█████╔╝ 
                          ╚════██║██╔═══╝ ██╔══╝  ██╔══██║██╔═██╗ 
                          ███████║██║     ███████╗██║  ██║██║  ██╗
                          ╚══════╝╚═╝     ╚══════╝╚═╝  ╚═╝╚═╝  ╚═╝

Talk to your Claude.

Voice cloning. Long documents. Audiobook quality. Local & private.

speak article.md --stream → Audio starts in seconds

---

Install

**For AI Agents** (Claude Code, Cursor, Windsurf):

npx skills add EmZod/speak

**CLI:**

git clone https://github.com/EmZod/speak.git
cd speak && bun install
alias speak="bun run $(pwd)/src/index.ts"

**Requirements:** macOS Apple Silicon · Bun · Python 3.10+ · sox (`brew install sox`)

---

Usage

speak "Hello, world!" --play        # Generate and play
speak article.md --stream           # Stream long content  
speak document.md --output out.wav  # Save to file
speak --clipboard --play            # Read from clipboard

---

Voice Cloning

Clone any voice from a 10-30 second sample:

# Use your cloned voice
speak "Hello" --voice ~/.chatter/voices/morgan_freeman.wav --play

---

Long Documents

speak book.md --auto-chunk --output book.wav    # Auto-chunk for reliability
speak --resume manifest.json                     # Resume interrupted generation
speak *.md --output-dir ~/Audio/                 # Batch processing
speak --estima