ElevenLabs
List voices, generate speech and sound effects, and transcribe bounded audio from a Swarmz project
Connect ElevenLabs when a project needs generated voice, sound effects, or speech transcription. The connector exposes explicit audio operations; it does not claim voice cloning or live streaming in this release.
ElevenLabs uses a shared project connection. Usage and provider billing belong to the connected ElevenLabs account.
Availability
A Swarmz administrator must show ElevenLabs under Platform config > Integrations. The workspace supplies its own API key; no Swarmz OAuth application is required.
Connect ElevenLabs
- In ElevenLabs, open your profile and create or copy an API key. Apply the narrowest product permissions and credit limit that fit the project. ElevenLabs documents API-key authentication.
- In the Swarmz project, open Settings > Integrations and select ElevenLabs.
- Select Add connection, paste the key into API key, and select Save connection.
Swarmz validates the key, encrypts it, and sends it only from the connector gateway to ElevenLabs.
Available operations
| Operation | What it does | Limit per minute |
|---|---|---|
list_voices | Lists voices available to the connected account | 30 |
text_to_speech | Generates one bounded MP3 from text and a voice ID | 10 |
sound_effect | Generates one bounded MP3 sound effect | 5 |
speech_to_text | Transcribes one bounded base64 audio file with Scribe v2 | 5 |
Audio generation and transcription are writes because they consume provider capacity. The AI asks for allow-once approval before each call. Generated audio returns as bounded base64 with its content type, ready for server-side storage or download handling.
Ask the AI to use ElevenLabs
- “List our available voices and build a preview picker before generating anything.”
- “Generate an MP3 reading this release note with voice ID
….” - “Create a two-second notification chime that does not loop.”
- “Transcribe this short audio upload and separate speakers where possible.”
For a published app, keep the gateway call in a server route. Store the returned audio in project storage when users need a durable URL; do not put the provider key in browser code.
Verify ElevenLabs
- List voices and confirm one known voice ID.
- Generate a short test sentence and play the returned MP3.
- Generate a short sound effect and confirm its duration and loop setting.
- Transcribe a small audio sample with a known phrase and compare the text.
- Check the ElevenLabs usage page for the matching requests.
Troubleshoot and recover
If a key validates but an operation is denied, review that key's product permissions and account quota. If audio is rejected as too large, shorten or compress it before retrying; the gateway deliberately bounds both uploads and responses.
To rotate the key, create a replacement in ElevenLabs, update the Swarmz connection, and verify list_voices before revoking the old key. To disconnect, select Remove in Swarmz and revoke the key in ElevenLabs.