Hello,
I have a question regarding my voice agent.
During conversations, the agent frequently uses a specific word that I would prefer it never says. I would like to know if there is a way to completely block or omit a specific word from the agent’s responses.
For example, I would like to configure the agent so that if the model generates that word, it is automatically removed or replaced before being spoken.
Is there any recommended way to achieve this? For example through:
-
system prompt rules
-
response filtering
-
post-processing before TTS
-
any built-in feature in Retell
My goal is simply to ensure that this word is never spoken during the conversation.
Thank you in advance for your help.