#speech-to-text #text-to-speech #voxtral #transcription

Voxtral Speech

@sliekens/voxtral

Install

openclaw plugins install clawhub:@sliekens/voxtral

@sliekens/voxtral

OpenClaw plugin that adds Voxtral text-to-speech and real-time speech-to-text via the Mistral API.

Capability	Config section	Provider ID	Default model
Speech (TTS)	`messages.tts`	`voxtral`	`voxtral-mini-tts-2603`
Realtime transcription (STT)	`plugins.entries.voice-call.config.streaming`	`voxtral`	`voxtral-mini-transcribe-realtime-2602`

Installation

bash

openclaw plugins install @sliekens/voxtral

API key

The plugin uses your Mistral API key. There are three ways to provide it, in order of priority:

1. Already have the Mistral core plugin configured?

Nothing to do. The Voxtral plugin will automatically pick up the Mistral credentials you already have configured.

2. Environment variable

bash

export MISTRAL_API_KEY=your-key-here

3. Explicit per-provider key in `openclaw.json`

json

{
  "messages": {
    "tts": {
      "providers": {
        "voxtral": {
          "apiKey": "your-key-here"
        }
      }
    }
  },
  "plugins": {
    "entries": {
      "voice-call": {
        "config": {
          "streaming": {
            "providers": {
              "voxtral": {
                "apiKey": "your-key-here"
              }
            }
          }
        }
      }
    }
  }
}

Configuration

Enable Voxtral TTS

json

{
  "messages": {
    "tts": {
      "provider": "voxtral"
    }
  }
}

Set a default voice:

json

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

Enable Voxtral STT

json

{
  "plugins": {
    "entries": {
      "voice-call": {
        "config": {
          "streaming": {
            "provider": "voxtral"
          }
        }
      }
    }
  }
}

Full example

json

{
  "messages": {
    "tts": {
      "provider": "voxtral",
      "providers": {
        "voxtral": {
          "apiKey": "your-key-here",
          "voice": "your-voice-id",
          "model": "voxtral-mini-tts-2603"
        }
      }
    }
  },
  "plugins": {
    "entries": {
      "voice-call": {
        "config": {
          "streaming": {
            "provider": "voxtral",
            "providers": {
              "voxtral": {
                "apiKey": "your-key-here",
                "model": "voxtral-mini-transcribe-realtime-2602"
              }
            }
          }
        }
      }
    }
  }
}

Directive tokens

Override the voice or model for a specific message using inline directive tokens:

text

[voxtral_voice: your-voice-id] Hello, this uses a specific voice.
[voxtral_model: voxtral-mini-tts-2603] Hello, this uses a specific model.

Advanced options

Option	Where	Default	Description
`apiKey`	TTS + STT provider config	—	Mistral API key
`baseUrl`	TTS + STT provider config	`https://api.mistral.ai/v1`	Custom API endpoint
`model`	TTS + STT provider config	see above	Override the default model
`voice`	TTS provider config	—	Default voice ID
`sampleRate`	STT provider config	`8000`	Audio sample rate in Hz (e.g. `8000`, `16000`)
`silenceDurationMs`	STT provider config	`800`	Milliseconds of silence before flushing the current utterance

You can also set the base URL via environment variable:

bash

export MISTRAL_TTS_BASE_URL=https://your-endpoint/v1
export MISTRAL_STT_BASE_URL=https://your-endpoint/v1

Release notes

1.0.3

Fixes compatibility with newer OpenClaw versions
- By prebuilding index.ts to dist/index.js and including it in the ClawHub package.

1.0.2

Fix compatibility with newer OpenClaw versions
- By adding missing activation.onStartup field to openclaw.plugin.json

1.0.1

Fix compatibility with newer OpenClaw versions
- By replacing deprecated providerAuthEnvVars with setup.providers[].envVars
Add separate CHANGELOG.md file

1.0.0

Initial release
Supports Voxtral TTS and real-time STT via the Mistral API
Supports reusing the same Mistral API key as the core plugin, or specifying separate keys for TTS and STT
Supports voices and models (currently only one TTS and one STT model, but more may be added in the future)
Supports directive tokens to override the voice or model on a per-message basis
Configurable sample rate and silence duration for STT
Integrates with the OpenClaw messages and voice call plugins
Integrates with Twilio via voice call plugin for real-time transcription of phone calls

Voxtral Speech

Install

@sliekens/voxtral

Installation

API key

1. Already have the Mistral core plugin configured?

2. Environment variable

3. Explicit per-provider key in openclaw.json

Configuration

Enable Voxtral TTS

Enable Voxtral STT

Full example

Directive tokens

Advanced options

Release notes

1.0.3

1.0.2

1.0.1

1.0.0

3. Explicit per-provider key in `openclaw.json`