emzod-speak-turbo

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

Talk to your Claude.

~90ms to first sound. Realistic. Local. Private. Fast.

speakturbo "Hello world" → ⚡ 92ms → ▶ 93ms → ✓ done

---

Install

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

npx skills add EmZod/Speak-Turbo

CLI only:

pip install pocket-tts uvicorn fastapi
cd speakturbo-cli && cargo build --release

---

Usage

speakturbo "Hello world"              # Play instantly
speakturbo "Hello" -o out.wav         # Save to file
speakturbo "Hello" -q                 # Quiet mode
speakturbo --list-voices              # Show voices

---

Voices

alba      ██████████  Female (default)
marius    ██████████  Male
javert    ██████████  Male  
jean      ██████████  Male
fantine   ██████████  Female
cosette   ██████████  Female
eponine   ██████████  Female
azelma    ██████████  Female

---

Performance

Time to first sound    ░░░░░░░░░░░░░░░░░░░░  ~90ms
First run (cold)       ████░░░░░░░░░░░░░░░░  2-5s  
Real-time factor       ████████████████░░░░  4x faster

---

Architecture

                    ┌─────────────────┐
                    │   speakturbo    │
                    │   (Rust, 2.2MB) │
                    └────────┬────────┘
                             │ HTTP :7125
                             ▼
                    ┌─────────────────┐
                    │     daemon      │
                    │ (Python + MLX)  │
                    └────────┬────────┘
                             │
                             ▼
                    ┌─────────────────┐
                    │  Audio Output   │
                    │    (rodio)      │
                    └─────────────────┘

---

Troubleshooting

Problem	Fix
No audio	curl http://127.0.0.1:7125/health
Daemon stuck	pkill -f "daemon_streaming"
Slow first run	Normal - model loading (2-5s)

---

See Also

Need voice cloning? Emotion tags? Try speak.

---

_{MIT License · Built on Pocket TTS}