Integrations · Voice & language AI
ElevenLabs on VegaDūta: voice for your agents
VegaDūta integrates ElevenLabs for agent voice replies: multilingual text-to-speech turns an agent's answer into speech, and instant voice cloning lets you create custom voices stored in a per-tenant voice library. Usage is governed by quotas, and the capability is currently in beta. To be precise about scope: VegaDūta uses ElevenLabs for TTS and voice cloning only — it does not use ElevenLabs' Conversational AI agents. The conversation intelligence is VegaDūta's own agent runtime; ElevenLabs supplies the voice.
What the integration provides
Two ElevenLabs capabilities are wired in, both in service of giving your agent a voice on channels where spoken replies make sense.
- Text-to-speech with ElevenLabs' multilingual model for agent voice replies.
- Instant voice cloning — create a custom voice and reuse it.
- A per-tenant voice library, isolated per workspace.
- Usage quotas on voice generation.
Honest scope: TTS and cloning, not ElevenLabs agents
ElevenLabs also sells a Conversational AI agent product; VegaDūta does not use it. Your agent's reasoning, tools, guardrails, and memory are all VegaDūta's runtime — ElevenLabs is the speech synthesis layer on top. If you specifically want ElevenLabs' own conversational agents, this integration is not that, and we would rather say so here.
Where agent voice fits
Voice replies matter on channels where users speak rather than type, and for accessibility. Paired with the Sarvam integration's speech-to-text, an agent can hear a voice message, reason over it, and answer aloud — a full voice loop with each layer handled by the provider best at it. The capability is in beta, so expect it to keep evolving.
Frequently asked questions
Can my AI agent reply with a voice?
Yes. VegaDūta uses ElevenLabs' multilingual text-to-speech to turn agent replies into speech, using a voice from your tenant's voice library. The capability is currently in beta.
Can I clone a voice for my agent?
Yes. The integration supports ElevenLabs instant voice cloning: you create a custom voice, it is stored in your per-tenant voice library, and your agents can speak with it. Voice generation is governed by quotas.
Does VegaDūta use ElevenLabs Conversational AI agents?
No. VegaDūta uses ElevenLabs for text-to-speech and voice cloning only. The conversational intelligence — reasoning, tools, guardrails — is VegaDūta's own agent runtime, with ElevenLabs providing the voice output.
Are cloned voices shared between customers?
No. Voices live in a per-tenant voice library, consistent with the platform's schema-per-tenant isolation — your workspace's voices are yours alone.
See it working in two minutes
The sandbox provisions a real tenant — describe an agent in one sentence and test it, no account, no card. Or browse ~90 industry workflow recipes to see what teams build.
