Inworld TTS for OpenClaw

Unofficial, community-maintained OpenClaw plugin that adds the Inworld TTS provider. It is built for OpenClaw's extension loader and ClawHub publication, not as a generic Inworld SDK wrapper.

• Package name: openclaw-inworld-tts
• Extension/plugin id: inworld-tts
• Speech provider id: inworld
• License: MIT
• Required auth: INWORLD_API_KEY or messages.tts.providers.inworld.apiKey

The extension registers the inworld speech provider, so OpenClaw config still uses messages.tts.provider: "inworld" and messages.tts.providers.inworld.

Requirements

• OpenClaw >=2026.4.1
• Inworld API key via INWORLD_API_KEY or messages.tts.providers.inworld.apiKey
• An Inworld voiceId

Install

Once published, install it like any other external OpenClaw plugin:

openclaw plugins install openclaw-inworld-tts

OpenClaw checks ClawHub first for bare npm-safe package names and falls back to npm if needed.

Minimal TTS config

{
  messages: {
    tts: {
      provider: "inworld",
      providers: {
        inworld: {
          voiceId: "your-inworld-voice-id",
        },
      },
    },
  },
}

Example with all supported settings

{
  messages: {
    tts: {
      provider: "inworld",
      providers: {
        inworld: {
          apiKey: "optional-if-INWORLD_API_KEY-is-set",
          baseUrl: "https://api.inworld.ai",
          voiceId: "your-inworld-voice-id",
          modelId: "inworld-tts-1.5-max",
          temperature: 0.7,
          timestampType: "WORD",
          applyTextNormalization: "auto",
          languages: ["ja-JP", "en-US"],
          audioConfig: {
            audioEncoding: "MP3",
            sampleRateHertz: 44100,
            bitRate: 128000,
            speakingRate: 1,
          },
        },
      },
    },
  },
}

Talk mode example

Talk mode uses talk.provider and talk.providers.<provider>. For Inworld, keep the Talk provider name as inworld.

{
  talk: {
    provider: "inworld",
    providers: {
      inworld: {
        apiKey: "optional-if-INWORLD_API_KEY-is-set",
        baseUrl: "https://api.inworld.ai",
        voiceId: "your-inworld-voice-id",
        modelId: "inworld-tts-1.5-max",
        temperature: 0.7,
        applyTextNormalization: "auto",
        audioConfig: {
          audioEncoding: "MP3",
          sampleRateHertz: 44100,
          bitRate: 128000,
          speakingRate: 1,
        },
      },
    },
  },
}

talk.speak override example

The public talk.speak gateway method accepts the generic Talk override fields. For the Inworld provider, the useful per-request overrides are voiceId, modelId, speed, and normalize.

{
  "method": "talk.speak",
  "params": {
    "text": "こんにちは、OpenClawです。",
    "voiceId": "alternate-voice-id",
    "modelId": "inworld-tts-1.5-max",
    "speed": 0.95,
    "normalize": "on"
  }
}

Provider-specific defaults still belong in talk.providers.inworld.

Discord voice example

Discord voice playback can override the global messages.tts config with channels.discord.voice.tts.

{
  channels: {
    discord: {
      voice: {
        enabled: true,
        tts: {
          provider: "inworld",
          providers: {
            inworld: {
              voiceId: "your-inworld-voice-id",
              modelId: "inworld-tts-1.5-max",
              audioConfig: {
                audioEncoding: "OGG_OPUS",
                sampleRateHertz: 48000,
              },
            },
          },
        },
      },
    },
  },
}

You can also override per Discord account with channels.discord.accounts.<account-id>.voice.tts when you want one bot account to speak differently from the global Discord voice default.

Config reference

Supported values

modelId

• inworld-tts-1
• inworld-tts-1-max
• inworld-tts-1.5-mini
• inworld-tts-1.5-max

audioConfig.audioEncoding

• LINEAR16
• MP3
• OGG_OPUS
• ALAW
• MULAW
• FLAC
• PCM
• WAV

audioConfig.sampleRateHertz

• 8000
• 16000
• 22050
• 24000
• 32000
• 44100
• 48000

Target-specific defaults

When audioConfig is omitted, OpenClaw picks defaults based on the output target.

OpenClaw behavior notes

• The extension registers only a speech provider. It does not add a new channel or realtime voice transport by itself.
• messages.tts.provider must remain inworld. The plugin name inworld-tts is only for OpenClaw's plugin loader.
• messages.tts.providers.inworld.languages affects listVoices, not the synthesis request.
• talk.providers.inworld can store provider-specific Talk defaults such as baseUrl, temperature, applyTextNormalization, and audioConfig.
• Public talk.speak requests use the generic Talk schema; for Inworld the practical per-request overrides are voiceId, modelId, speed, and normalize.
• channels.discord.voice.tts overrides messages.tts for Discord voice playback only.
• Telephony synthesis forces LINEAR16 at 8000 Hz.

Service docs

• Inworld API introduction
• Inworld TTS synthesize speech
• Inworld list voices

ClawHub packaging notes

• package.json includes openclaw.compat and openclaw.build, which are required for ClawHub-published external plugins.
• package.json keeps openclaw as a peer dependency because the host runtime provides it.
• The local test setup uses lightweight SDK shims so pnpm test and pnpm check can run without a sibling OpenClaw checkout.

Validation

From this package directory, first install dependencies so pnpm-lock.yaml is generated locally:

pnpm install

Then run:

pnpm test
pnpm check

• pnpm test runs the local extension Vitest suite, including readme.test.ts.
• pnpm check runs the local formatting check, local lint, and then the local extension test suite.

Support

If this plugin is useful, you can support development here:

• Sponsor on GitHub