elevenlabs-stt

Add ElevenLabs Speech-to-Text to OpenClaw's normal inbound audio pipeline (tools.media.audio).

The plugin is intentionally simple to use: install it, add an ElevenLabs provider block, and point an audio model entry at provider: elevenlabs. It also exposes selected ElevenLabs request options through providerOptions.elevenlabs, so you can tune transcript quality and behavior without leaving normal OpenClaw config.

Table of contents

• Prerequisites
• Quick start
• Configuration
• Provider options
• What this plugin covers
• Compatibility
• Development
• Documentation
• License

Prerequisites

• OpenClaw >=2026.3.24
• Node.js >=22.14.0
• An ElevenLabs API key

Quick start

Install with OpenClaw:

openclaw plugins install @mkv21/elevenlabs-stt

Use the scoped package name for OpenClaw installs. On current OpenClaw versions, the unscoped name elevenlabs-stt collides with an existing ClawHub skill during resolver lookup.

Warning Before restarting the OpenClaw Gateway after installation, add the full models.providers.elevenlabs config block shown below. If the plugin is enabled but the provider block is missing or incomplete, OpenClaw can fail validation during startup and the Gateway may no longer start.

Minimal working config:

plugins:
  entries:
    elevenlabs-stt:
      enabled: true

models:
  providers:
    elevenlabs:
      baseUrl: https://api.elevenlabs.io/v1
      apiKey: ${ELEVENLABS_API_KEY}
      models: []

tools:
  media:
    audio:
      enabled: true
      timeoutSeconds: 90
      language: en
      models:
        - provider: elevenlabs
          model: scribe_v2

Important notes:

• models.providers.elevenlabs.models: [] is currently required by OpenClaw's shared provider config shape.

• If you prefer an explicit env-backed secret ref, you can use:

  apiKey:
    source: env
    provider: default
    id: ELEVENLABS_API_KEY
  

• If you want fallback behavior, add additional entries after the ElevenLabs model in tools.media.audio.models.

Configuration

Provider options

Tune the upstream ElevenLabs request through providerOptions.elevenlabs, set globally on tools.media.audio.providerOptions.elevenlabs or per model entry on tools.media.audio.models[].providerOptions.elevenlabs.

Example:

tools:
  media:
    audio:
      providerOptions:
        elevenlabs:
          no_verbatim: true
          tag_audio_events: true
      models:
        - provider: elevenlabs
          model: scribe_v2

Note: OpenClaw currently consumes only the plain transcript text. Options whose main value is extra metadata (timestamps, speakers, channels, entities) are forwarded to ElevenLabs but not yet surfaced through OpenClaw's current audio result shape. Array-shaped options such as keyterms and additional_formats are not supported because current OpenClaw media providerOptions are scalar-only.

What this plugin covers

• Normal inbound audio transcription only
• No telephony, realtime/streaming, or async webhook-based transcription flow

Compatibility

• Tested locally with OpenClaw 2026.3.24
• Requires OpenClaw host version >=2026.3.24

Development

# Run tests
npm test

# Type check
npm run typecheck

Documentation

• Configuration Reference and Troubleshooting

AI-Assisted Development Disclaimer

Parts of this project were developed with assistance from generative AI tools. All generated code and text were reviewed and adapted by a human before publication.

License

This project is licensed under the MIT License. See LICENSE for details.