dsh-multi-tts
语音朗读(MiniMax / OpenAI 兼容 TTS,服务商与音色自选)——每条回复旁的朗读按钮、输入框自动朗读开关、设置页自选服务商与音色。零构建,纯 JS。
插件会安装到这里；不确定时保持 web。
npx -y @deepseek-ai/dsh plugin --profile web add github:WYR-233/dsh-multi-tts#7626a4f33693e8518549f42b315273079c412e6eMulti Tts 以 dsh-multi-tts 发布，当前版本为 0.1.4。Plugin Hub 会校验它的 manifest，并保存精确安装来源，便于复现安装结果。
Read assistant replies aloud in DeepSeek Harness — pick your own TTS provider and voice.
English | 中文
A multi-provider text-to-speech plugin for DSH: a 🔊 button next to every assistant reply, an auto-read toggle in the composer, and a settings page where you choose the provider, the voice, the emotion and the speed. No build step — plain ESM + hand-written React.
$DSH_HOME/.env → common key names in .env.React.createElement shipped as-is in lib/.| Provider | Endpoint | Built-in voices | Emotion |
|---|---|---|---|
minimax | MiniMax (Hailuo) /v1/t2a_v2 | 12 Chinese voices (e.g. female-shaonv, qiaopi_mengmei, lovely_girl) | ✅ |
openai | any OpenAI-compatible /audio/speech | 6 voices (alloy, echo, fable, onyx, nova, shimmer) | — |
Any OpenAI-compatible endpoint works through the openai provider — just set the base URL (OpenAI, DashScope, SiliconFlow, local servers, …).
dsh plugin --profile web add dsh-multi-tts
Then add the package to the profile's bundle list (dsh plugin add only writes the dependency — dsh.profile.bundles is what actually mounts the plugin):
{
"dsh": {
"profile": {
"bundles": ["dsh-multi-tts"]
}
}
}
Restart DSH, then open Settings → Multi TTS / 语音朗读.
$DSH_HOME/.env:MINIMAX_API_KEY=sk-...
OPENAI_API_KEY=sk-...
(from env/.env) when the key comes from the environment.Config is stored in $DSH_HOME/multi-tts/config.json (override the directory with the stateDir option).
The host half registers four routes on the DSH web server:
| Method | Path | Purpose |
|---|---|---|
POST | /multi-tts/synthesize | { text, voice?, provider?, speed? } → audio bytes (char count in x-tts-chars) |
GET | /multi-tts/status | active provider / voice / key presence |
GET POST | /multi-tts/config | read / patch the stored config |
GET | /multi-tts/voices | voice list of the active provider |
@deepseek-ai/cordis ^4.0.1, provided by the DSH host.fetch).MIT
在 DeepSeek Harness 里把助手的回复念出来——服务商与音色都由你自己挑。
DSH 的多服务商语音朗读插件：每条回复旁有 🔊 朗读按钮，输入框有自动朗读开关，设置页里可选服务商、音色、情绪与语速。零构建，纯 ESM + 手写 React。
$DSH_HOME/.env → .env 常见键名。React.createElement，直接以 lib/ 里的源文件形式分发。| 服务商 | 接口 | 内置音色 | 情绪 |
|---|---|---|---|
minimax | MiniMax(海螺)/v1/t2a_v2 | 12 个中文音色(female-shaonv 少女音、qiaopi_mengmei 俏皮萌妹、lovely_girl 萌萌女童……) | ✅ |
openai | 任意 OpenAI 兼容 /audio/speech | 6 个音色(alloy、echo、fable、onyx、nova、shimmer) | — |
凡是 OpenAI 兼容的语音端点都能用 openai 这条通道接进来——改 Base URL 即可(OpenAI、DashScope、SiliconFlow、本地服务……)。
dsh plugin --profile web add dsh-multi-tts
再把包名加进 profile 的 bundles 清单(dsh plugin add 只写依赖,bundles 才是挂载清单):
{
"dsh": {
"profile": {
"bundles": ["dsh-multi-tts"]
}
}
}
重启 DSH,打开设置 → 语音朗读。
$DSH_HOME/.env:MINIMAX_API_KEY=sk-...
OPENAI_API_KEY=sk-...
配置存在 $DSH_HOME/multi-tts/config.json(可用 stateDir 选项改目录)。
host 端在 DSH web 服务器上挂 4 条路由:
| 方法 | 路径 | 用途 |
|---|---|---|
POST | /multi-tts/synthesize | { text, voice?, provider?, speed? } → 音频字节(字符数在 x-tts-chars) |
GET | /multi-tts/status | 当前服务商 / 音色 / 是否已配 Key |
GET POST | /multi-tts/config | 读 / 改配置 |
GET | /multi-tts/voices | 当前服务商的音色表 |
@deepseek-ai/cordis ^4.0.1,由 DSH 本体提供。fetch)。MIT