Hi - can someone help me with listing ASR/TTS APIs on livekit for developers to use the same for building agents.

Thanks for reaching out, ankur.agrawal. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.

both

Thanks — below is a combined, concise list of ASR (STT) and TTS options on LiveKit including both LiveKit Inference (hosted) models and plugin integrations (bring-your-own keys), plus brief usage notes. ASR / Speech-to-Text (STT) • LiveKit Inference (hosted, no provider keys required) ◦ Deepgram: Flux, Nova-3 (45 languages), Nova-2 ◦ AssemblyAI: Universal-3 Pro Streaming, Universal-Streaming, Universal-Streaming-Multilingual ◦ Cartesia: Ink Whisper (100 languages) ◦ ElevenLabs: Scribe v2 Realtime (190 languages) ◦ Speechmatics: Enhanced and Standard (61 languages) ◦ xAI: Speech to Text (25 languages) ◦ (Usage) Configure via model ID, e.g.

inference.STT(model="assemblyai/u3-rt-pro", language="en")

• Plugins (use your provider API keys) ◦ Providers with plugin support: Deepgram, AssemblyAI, Azure AI Speech, Google Cloud, Amazon Transcribe, Cartesia, Clova, ElevenLabs, NVIDIA Riva, Web/Baseten/Gladia, and others. ◦ (Usage) Set provider keys in env (e.g.,

DEEPGRAM_API_KEY

) and instantiate plugin STT (example:

from livekit.plugins import deepgram; stt = deepgram.STT(model="nova-3", language="en")

) TTS / Text-to-Speech • LiveKit Inference (hosted, no provider keys required) ◦ Deepgram: Aura-2 ◦ ElevenLabs: Eleven Flash v2.5, Eleven Turbo v2.5, Eleven Multilingual v2 ◦ Cartesia: Sonic-3, Sonic Turbo ◦ Rime: Arcana, Mist series ◦ xAI: TTS-1 ◦ Inworld, Resemble AI, and others ◦ (Usage) Configure via model ID, e.g.

inference.TTS(model="cartesia/sonic-3", voice="VOICE_ID", language="en")

• Plugins (use your provider API keys) ◦ Providers with plugin support: Deepgram, ElevenLabs, Cartesia, Rime, Azure, Google, Amazon Polly, OpenAI, xAI, and others. ◦ (Usage) Set provider keys in env (e.g.,

ELEVEN_API_KEY

) and instantiate plugin TTS (example:

from livekit.plugins import elevenlabs; tts = elevenlabs.TTS(voice_id="...", model="eleven_multilingual_v2")

) Additional notes • Both STT and TTS can be used inside

AgentSession

or standalone. • LiveKit supports shorthand string descriptors for models (e.g.,

tts="xai/tts-1:ara"

). • Features available across providers: speaker diarization, turn detection, expressive speech tags, and keyterm boosting (provider-dependent). • For plugin usage, set provider API keys in your environment (e.g.,

.env

) before creating plugin instances. If you’d like, I can: • Produce a sortable table (hosted vs plugin) with model names, languages supported, and sample code snippets for Python or Node.js. • Provide step-by-step examples for a specific provider (pick one) showing env vars, initialization, and a minimal AgentSession example. Which follow-up would you prefer? , with interactive elements

Is Sarvam supported?

Yes — Sarvam is supported on LiveKit via plugins for both STT (ASR) and TTS. You’ll need to provide your Sarvam API keys (set them in your environment) and instantiate the Sarvam plugin when creating STT/TTS instances. Would you like a short Python example for initializing Sarvam STT or TTS? Sources: Speech-to-text (STT) models overview | LiveKit Documentation | LiveKit Inference | LiveKit Documentation | Text-to-speech (TTS) models overview | LiveKit Documentation, with interactive elements