say

在 macOS 上调用系统 say 命令进行本地文本转语音，可用 Siri 自然语音或通过 -v 指定非 Siri 语音，用 defaults write 切换 Siri 语言，支持设置语速、输出 AIFF 并用 ffmpeg 转换为 WAV。

tobihagemann

热度 78.72

最近更新 2026-05-11 16:52:13

分类：创作依赖：代码执行 / 读写文件应用场景：工作流 / 文件转换

安装技能

复制下面的话术发给你的 Agent，它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill： npx secure-skills-cli add say

Metadata

name: say
description: Text-to-Speech via macOS say command with Siri Natural Voices. Use for generating speech audio, TTS clips, or speaking text aloud on macOS.
metadata: {"openclaw":{"emoji":"🔊","os":["darwin"],"requires":{"bins":["say","ffmpeg"]}}}

say

Use say for on-device text-to-speech on macOS.

Siri Natural Voices (recommended)

Siri voices are the best macOS TTS voices but cannot be selected via -v. Instead, run say without -v — it uses the system default voice. Switch languages via defaults write:

# Switch to German
defaults write com.apple.speech.voice.prefs SystemTTSLanguage -string "de"
say "Hallo, wie geht's?" -o output_de.aiff

# Switch to Chinese (Mandarin)
defaults write com.apple.speech.voice.prefs SystemTTSLanguage -string "cmn"
say "你好，世界" -o output_zh.aiff

No process restart needed — the next say invocation picks up the new language immediately.

Prerequisites

Download the desired Siri voices first in System Settings > Accessibility > Spoken Content and set them as the system voice for each language.

Check which voices are currently configured:

defaults read com.apple.Accessibility SpokenContentDefaultVoiceSelectionsByLanguage

Fallback: select voice via `-v`

For non-Siri voices, use -v directly:

say -v 'Tingting (Enhanced)' "你好，世界"
say -v '?'  # list all installed voices (Siri voices not listed)

Output to file

say -o output.aiff "Hello world"
ffmpeg -y -i output.aiff -ar 22050 -ac 1 output.wav  # convert to WAV

Options

-v <voice> — Select a non-Siri voice
-r <rate> — Speaking rate in words per minute (e.g. -r 150)
-o <file> — Save to AIFF file instead of playing aloud

Notes

say adds natural pauses at punctuation — no manual sentence splitting needed
AIFF is the native output format; convert with ffmpeg for WAV/MP3
For batch generation: set language once, generate all clips, then switch — minimizes defaults write calls

say