### Answer Based on the available sources, this issue with "ums" appearing in agent responses has been reported by other users. There's a Backchanneling setting in Retell AI that should control this behavior. [(1) ]() One user experienced a sudden increase in "Um"s being injected into responses even though they had Backchanneling turned off. The utterances appeared in the transcript, suggesting it wasn't a TTS issue but rather coming from the LLM itself. The recommended solution is to explicitly prompt your agent to never use filler words in your system prompt. While you shouldn't need to do this if Backchanneling is disabled, adding this instruction to your prompt has been confirmed to resolve the issue. Check your agent's Backchanneling setting to ensure it's turned off, and if the issue persists, add explicit instructions in your prompt telling the agent not to use filler words like "um" or "uh".

​ To learn more, see the sources I considered: [[1] Huge increase backchanneling "Um"s](https://discord.com/channels/1186424376300679209/1438582091905499239) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.