Thanks — I’m happy to help you get LiveKit TTS working with Deepgram’s Aura‑2. I’ll cover the quickest ways to get started and ask one quick question so I can tailor the answer to your setup. Quick question before we dive in: • Do you want to use LiveKit’s built‑in Deepgram support (no Deepgram API key needed, billed via LiveKit) or do you plan to use your own Deepgram API key via the Deepgram plugin? If you want the built‑in LiveKit option (recommended for simplicity) • You can specify the Deepgram Aura‑2 model directly when creating an AgentSession. • Minimal examples: 1) Python (string descriptor)

from livekit.agents import AgentSession

session = AgentSession(
    tts="deepgram/aura-2:athena",
    # add llm, stt, etc. as needed
)

2) TypeScript (string descriptor)

import { AgentSession } from '@livekit/agents';

const session = new AgentSession({
  tts: "deepgram/aura-2:athena",
  // add llm, stt, etc.
});

• Use the

model: "deepgram/aura-2"

form if you need more control or want to set language/extra options:

from livekit.agents import AgentSession, inference

session = AgentSession(
  tts=inference.TTS(
    model="deepgram/aura-2",
    voice="athena",
    language="en"
  )
)

Notes on voices and options • Supported Aura voices:

athena

,

apollo

,

odysseus

,

theia

. •

language

is optional and defaults to the model default. • You can pass

extra_kwargs

(or

modelOptions

in Node) to send additional Deepgram TTS parameters if you need special audio tuning. If you want to use your own Deepgram API key (plugin approach) • Install the plugin: ◦ Python:

uv add "livekit-agents[deepgram]~=1.4"

◦ Node:

pnpm add @livekit/agents-plugin-deepgram@1.x

• Set

DEEPGRAM_API_KEY

in your environment. • Example using the plugin:

from livekit.plugins import deepgram
from livekit.agents import AgentSession

session = AgentSession(
  tts=deepgram.TTS(model="aura-2-asteria-en"),
  # ...
)

• The plugin’s default model is

aura-2-asteria-en

. Pricing and docs • LiveKit bills for Deepgram TTS through LiveKit Inference (Aura‑2 pricing listed on LiveKit pricing pages). • For Deepgram model/voice and advanced options, check Deepgram’s TTS docs: https://developers.deepgram.com/docs/tts-models Next steps / how I can help further • Tell me which option you prefer (LiveKit built‑in or your own Deepgram key). • Share the language, voice, or any audio parameters you want (e.g., rate, pauses, SSML-like behavior). • If you already tried code that’s failing, paste the snippet and the error and I’ll troubleshoot it with you. Which route would you like to take? Sources: Deepgram TTS | LiveKit Documentation | Build voice, video, and physical AI | LiveKit, with interactive elements