What is the best way to inject filler words?
# ask-aiq
quick-artist-10782
03/26/2025, 3:56 AMWhat is the best way to inject filler words?
t
tall-airline-16756
03/26/2025, 3:56 AMThanks for reaching out, Zach. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.
tall-airline-16756
03/26/2025, 3:56 AMI'm sorry, but I don't have the information you need at the moment. If you would like a member of the LiveKit team to respond, please ask your question in one of the other channels.
h
helpful-teacher-46454
03/26/2025, 5:33 AMmaybe you can use the agent.say method try working with it
helpful-teacher-46454
03/26/2025, 5:36 AMif you are trying to add filler words to reduce the latency of the bots response you can add a before llm callback and set some filler words over there
👍 1
a
acoustic-guitar-10279
03/29/2025, 8:58 AM@quick-artist-10782 @helpful-teacher-46454 do you know any better ways to improve latency other than hosting the agent? any changes in voicepipeline or any other way?
currently I'm getting avg latency of 1.7
h
helpful-teacher-46454
03/29/2025, 12:45 PM@acoustic-guitar-10279 could you provide a bit more insight of how your current pipeline is configured and exact use case
a
acoustic-guitar-10279
03/29/2025, 4:17 PMIm trying build a voicepipeline agent for outbound calling. Stt using deepgram,tts using cartesia, llm using cerebras. main focus is to reduce latency
h
helpful-teacher-46454
03/30/2025, 12:38 AM@acoustic-guitar-10279 my advice would be use logging and find out how much time each of your service in the pipeline takes and use different techniques like chunking and streaming data immediately via deeparam, using prewarm for Cartesia, try async response ,etc.
Might even consider switch models that perform better based on your budget