Title: How to pass

language

param in TTS string descriptor for

AgentSession

? Question: Hi everyone, I'm setting up an

AgentSession

using LiveKit Inference. I know I can configure TTS using the shorthand string descriptor like this: session = AgentSession( # Format: provider/model:voice_id tts="elevenlabs/eleven_turbo_v2_5:Xb7hH8MSUJpSbSDYk0k2", # ... llm, stt, vad, etc. ) However, I also need to specify the

language

option (e.g., forcing

"en"

or another language code). I know I can achieve this by importing

inference

and instantiating the object explicitly: from livekit.agents import AgentSession, inference session = AgentSession( tts=inference.TTS( model="elevenlabs/eleven_turbo_v2_5", voice="Xb7hH8MSUJpSbSDYk0k2", language="en" # <--- This is what I need ), # ... ) Is there a way to pass this

language

option directly in the string descriptor format? (e.g., something like

"elevenlabs/eleven_turbo_v2_5:voice_id?language=en"

)? Or is the object-based instantiation the only way to handle parameters beyond just the model and voice ID? Thanks!

It is not supported in the TTS from_model_string today. So this is the method you will need to use.

it is not supported in the TTS from_model_string today. in future?

Good question. I am not sure on that