Hey guys trying to wrap my head around how to control speed of speech for Eleven labs voice (with other providers just passing speed to extra_kwargs worked but with Elevenlabs no). Tried passing it a few ways, and there is no working one. The one in code here commented did not work at all, agent just didn't connect to the session room. Can you please help? session = AgentSession( stt=inference.STT(model="assemblyai/universal-streaming", language="en"), llm=inference.LLM(model="openai/gpt-4.1-mini"), tts=inference.TTS( model="elevenlabs/eleven_turbo_v2_5", voice="iP95p4xoKVk53GoZ742B", language="en-AU", extra_kwargs={"voice_settings": {"speed": 0.9}}, ), # tts=elevenlabs.TTS( # voice_id="iP95p4xoKVk53GoZ742B", # language="en-AU", # model="eleven_turbo_v2_5", # voice_settings=elevenlabs.VoiceSettings( # speed=0.9, stability=0.71, similarity_boost=0.5 # ), # ), turn_detection=MultilingualModel(), vad=ctx.proc.userdata["vad"], preemptive_generation=True, ) (edited)

Hi, there's a couple of things here 1. Right now, we are still working on reaching feature parity between our inference and plugin implementations. As far as I can see, speed is not supported in inference today, which is why your uncommented code is not having an effect 2. The plugin should work, but I didn't have any luck with your specific settings (I'm not entirely sure why). This worked for me:

# tts=elevenlabs.TTS(
        #     voice_id="Xb7hH8MSUJpSbSDYk0k2",
        #     model="eleven_turbo_v2_5",
        #     voice_settings=elevenlabs.VoiceSettings(
        #         speed=1.2, stability=0.71, similarity_boost=0.5
        #     ),
        # ),