ElevenLabs Automation

When to use this skill

Use ElevenLabs Automation when an AI agent needs a reusable SKILL.md workflow for this job: Automate ElevenLabs text-to-speech workflows -- generate speech from text, browse and inspect voices, check subscription limits, list models, stream audio, and retrieve history via the Composio MCP integration.

When not to use it

Skip ElevenLabs Automation when the task is outside the creativity category, or when a more specific skill in this directory already covers the same workflow with clearer triggers.

How to install

1. Personal install: create ~/.claude/skills/elevenlabs-automation/SKILL.md (and any bundled scripts) so Claude Code, Claude Desktop, and compatible agents can load it in every project.
2. Project install: commit the same folder at .claude/skills/elevenlabs-automation/ so teammates get the skill with the repo.
3. Restart the agent session after copying files so it re-scans the skills directory, then ask for the task in words that match the skill description.

Full install guide for Claude, Cursor, and Codex

What this skill does

# ElevenLabs Automation

Automate your ElevenLabs text-to-speech workflows -- convert text to natural speech, browse the voice library, inspect voice details, check subscription credits, select TTS models, stream audio for low-latency delivery, and retrieve previously generated audio from history.

**Toolkit docs:** [composio.dev/toolkits/elevenlabs](https://composio.dev/toolkits/elevenlabs)

---

## Setup

1. Add the Composio MCP server to your client: `https://rube.app/mcp` 2. Connect your ElevenLabs account when prompted (API key authentication) 3. Start using the workflows below

---

## Core Workflows

### 1. Generate Speech from Text

Use `ELEVENLABS_TEXT_TO_SPEECH` to convert text into a downloadable audio file.

``` Tool: ELEVENLABS_TEXT_TO_SPEECH Inputs: - voice_id: string (required) -- obtain from ELEVENLABS_GET_VOICES - text: string (required) -- max ~10,000 chars (most models), 30,000 (Flash/Turbo v2), 40,000 (v2.5) - model_id: string (default "eleven_monolingual_v1") -- e.g., "eleven_multilingual_v2" - output_format: string (default "mp3_44100_128") -- see formats below - optimize_streaming_latency: integer (0-4; NOT supported with eleven_v3) - seed: integer (optional, for reproducibility -- not guaranteed) - pronunciation_dictionary_locators: array (optional, up to 3 dictionaries) ```

**Output formats:** - MP3: `mp3_22050_32`, `mp3_44100_32`, `mp3_44100_64`, `mp3_44100_96`, `mp3_44100_128`, `mp3_44100_192` (Creator+) - PCM: `pcm_16000`, `pcm_22050`, `pcm_24000`, `pcm_44100` (Pro+) - uLaw: `ulaw_8000` (for Twilio)

**Important:** Output is a file object with a presigned download link at `data.file.s3url` (expires in ~1 hour). Download promptly.

### 2. Browse Available Voices

Use `ELEVENLABS_GET_VOICES` to list all voices with their attributes and settings.

``` Tool: ELEVENLABS_GET_VOICES Inputs: (none) ```

Returns an array at `data.voices[]` with `voice_id`, `name`, `labels` (gender, accent, use_case), and settings.

### 3. Inspect a Specific Voice

Use `ELEVENLABS_GET_VOICE` to get detailed metadata for a candidate voice before synthesis.

``` Tool: ELEVENLABS_GET_VOICE Inputs: - voice_id: string (required) -- e.g., "21m00Tcm4TlvDq8ikWAM" - with_settings: boolean (default false) -- include detailed voice settings ```

### 4. Check Subscription and Credits

Use `ELEVENLABS_GET_USER_SUBSCRIPTION_INFO` to verify plan limits and remaining credits before bulk generation.

``` Tool: ELEVENLABS_GET_USER_SUBSCRIPTION_INFO Inputs: (none) ```

### 5. List Available TTS Models

Use `ELEVENLABS_GET_MODELS` to discover compatible models and filter by `can_do_text_to_speech: true`.

``` Tool: ELEVENLABS_GET_MODELS Inputs: (none) ```

### 6. Stream Audio and Retrieve History

Use `ELEVENLABS_TEXT_TO_SPEECH_STREAM` for low-latency streamed delivery, and `ELEVENLABS_GET_AUDIO_FROM_HISTORY_ITEM` to re-download previously generated audio.

``` Tool: ELEVENLABS_TEXT_TO_SPEECH_STREAM - Same core inputs as TEXT_TO_SPEECH but returns a stream for low-latency playback

Tool: ELEVENLABS_GET_AUDIO_FROM_HISTORY_ITEM - history_item_id: string (required) -- ID from a previous generation ```

---

## Known Pitfalls

| Pitfall | Detail | |---------|--------| | Text length limits | Most models cap at ~10,000-20,000 chars per request. Oversized input returns HTTP 400. Split long text into chunks (~5000 chars) and generate per chunk. | | Output is a presigned URL | `ELEVENLABS_TEXT_TO_SPEECH` returns `data.file.s3url` with a ~1 hour expiry (X-Amz-Expires=3600). Download the audio file promptly. | | Quota and credit errors | HTTP 401 with `quota_exceeded` or HTTP 402 `payment_required` means insufficient credits or tier restrictions. Check with `ELEVENLABS_GET_USER_SUBSCRIPTION_INFO` before bulk jobs. | | Voice permissions | HTTP 401 with `missing_permissions` means the API key lacks `voices_read` scope. Verify key permissions. | | Model compatibility | Not all models support TTS. Use `ELEVENLABS_GET_MODELS` and filter by `can_do_text_to_speech: true`. The `optimize_streaming_latency` parameter is NOT supported with `eleven_v3`. | | Large voice list truncation | `ELEVENLABS_GET_VOICES` may return a large list. Select from the full `data.voices[]` payload -- previews may appear truncated. |

---

## Quick Reference

| Tool Slug | Description | |-----------|-------------| | `ELEVENLABS_TEXT_TO_SPEECH` | Convert text to speech, returns downloadable audio file | | `ELEVENLABS_GET_VOICES` | List all available voices with attributes | | `ELEVENLABS_GET_VOICE` | Get detailed info for a specific voice | | `ELEVENLABS_GET_USER_SUBSCRIPTION_INFO` | Check subscription plan and remaining credits | | `ELEVENLABS_GET_MODELS` | List available TTS models and capabilities | | `ELEVENLABS_TEXT_TO_SPEECH_STREAM` | Stream audio for low-latency delivery | | `ELEVENLABS_GET_AUDIO_FROM_HISTORY_ITEM` | Re-download audio from generation history |

---

*Powered by [Composio](https://composio.dev)*

Intended uses

• Use ElevenLabs Automation when this documented workflow matches the task.