菜单
Give your agent the ability to speak to you real-time. Talk to your Claude! Ultra-fast TTS, text-to-speech, voice synthesis, audio output with ~90ms latency. 8 built-in voices for instant voice responses. For voice cloning, use the speak skill.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | speakturbo-tts |
| description | Give your agent the ability to speak to you real-time. Talk to your Claude! Ultra-fast TTS, text-to-speech, voice synthesis, audio output with ~90ms latency. 8 built-in voices for instant voice responses. For voice cloning, use the speak skill. |
Give your agent the ability to speak to you real-time. Ultra-fast text-to-speech with ~90ms latency and 8 built-in voices.
# Play immediately - you should hear "Hello world" through your speakers
speakturbo "Hello world"
# Output: ⚡ 92ms → ▶ 93ms → ✓ 1245ms
# Verify it's working by saving to file
speakturbo "Hello world" -o test.wav
ls -lh test.wav # Should show ~50-100KB file
Output explained: ⚡ = first audio received, ▶ = playback started, ✓ = done
The first execution takes 2-5 seconds while the daemon starts and loads the model into memory. Subsequent calls are ~90ms to first sound.
# First run (slow - daemon starting)
speakturbo "Starting up" # ~2-5 seconds
# Second run (fast - daemon already running)
speakturbo "Now I'm fast" # ~90ms
# Basic - plays immediately (default voice: alba)
speakturbo "Hello world"
# Save to file (no audio playback)
speakturbo "Hello" -o output.wav
# Save to specific file
speakturbo "Goodbye" -o goodbye.wav
# Quiet mode (suppress status messages, still plays audio)
speakturbo "Hello" -q
# List available voices
speakturbo --list-voices
| Voice | Type |
|---|---|
alba | Female (default) |
marius | Male |
javert | Male |
jean | Male |
fantine | Female |
cosette | Female |
eponine | Female |
azelma | Female |
| Metric | Value |
|---|---|
| Time to first sound | ~90ms (daemon warm) |
| First run | 2-5s (daemon startup) |
| Real-time factor | ~4x faster |
| Sample rate | 24kHz mono |
speakturbo (Rust CLI, 2.2MB)
│
│ HTTP streaming (port 7125)
▼
speakturbo-daemon (Python + pocket-tts)
│
│ Model in memory, auto-shutdown after 1hr idle
▼
Audio playback (rodio)
speakturbo "She said \"hello\""The -o flag only writes to directories that are on the allowlist. By default, these are:
/tmp and system temp directories~/.speakturbo/If you need to write elsewhere, use --allow-dir:
speakturbo "Hello" -o /custom/path/audio.wav --allow-dir /custom/path
To permanently allow a directory, add it to ~/.speakturbo/config:
mkdir -p ~/.speakturbo && echo "/custom/path" >> ~/.speakturbo/config
The config file is one directory per line. Lines starting with # are comments.
| Code | Meaning |
|---|---|
| 0 | Success (audio played/saved) |
| 1 | Error (daemon connection failed, invalid args) |
Use speakturbo when:
Use speak instead when:
speak "text" --voice ~/.chatter/voices/morgan_freeman.wav[laugh], [sigh]See the speak skill documentation for full usage.
No audio plays:
# Check daemon is running
curl http://127.0.0.1:7125/health
# Expected: {"status":"ready","voices":["alba","marius",...]}
# Verify by saving to file and playing manually
speakturbo "test" -o /tmp/test.wav
afplay /tmp/test.wav # macOS
aplay /tmp/test.wav # Linux
Daemon won't start:
# Check port availability
lsof -i :7125
# Manually kill and restart
pkill -f "daemon_streaming"
speakturbo "test" # Auto-restarts daemon
First run is slow: This is expected. The daemon needs to load the ~100MB model into memory. Subsequent calls will be fast (~90ms).
The daemon auto-starts on first use and auto-shuts down after 1 hour idle.
# Check status
curl http://127.0.0.1:7125/health
# Manual stop
pkill -f "daemon_streaming"
# View logs
cat /tmp/speakturbo.log
| Feature | speakturbo | speak |
|---|---|---|
| Time to first sound | ~90ms | ~4-8s |
| Voice cloning | ❌ | ✅ |
| Emotion tags | ❌ | ✅ |
| Voices | 8 built-in | Custom wav files |
| Engine | pocket-tts | Chatterbox |
Voice-driven code agent. Talk to your codebase in natural language with speech-friendly responses. Triggers on "codechat", "voice chat about code", "talk to my codebase", or "chat about code".
Use when the user wants the KADE plus gstack bridge, including session kickoff with user-profile context, /g-kade install style project scaffolding, repo adaptation, or structured handoff logging tied to llm-wiki-memory packet setup. This packet-owned wrapper is installable into ~/.agents/skills, ~/.codex/skills, and ~/.claude/skills.
Use when the user wants gstack-style execution workflows such as QA, browser dogfooding, code review, debugging, ship or PR preparation, design review, benchmark checks, or deployment verification. This packet-owned wrapper is installable into ~/.agents/skills, ~/.codex/skills, and ~/.claude/skills.
Create distinctive, production-grade interfaces with high design quality using the Impeccable frontend vocabulary. Use when designing polished web or Android app experiences, especially when the user wants stronger visual taste, clearer UX, or a more memorable aesthetic.
Use when the user wants the KADE System itself, including Layer 1 and Layer 2 context loading, KADE session structure, HUMAN.md and kade/ overlays, or project initialization and handoff workflows. This packet-owned wrapper is installable into ~/.agents/skills, ~/.codex/skills, and ~/.claude/skills.
Sync agent skills from a canonical source directory into one or more Obsidian vaults or project workspaces. Compares skill directories by file count and total size, reports new/updated/vault-only skills, then copies everything from canonical into the target while preserving target-only skills. Triggers on "sync skills", "update skills", "skill sync", "refresh skills", "pull latest skills", "sync vault skills", "check for skill updates", or "are my skills up to date".