First, I want to thank you for this amazing project. 2001 is my favorite movie of all time. HAL is obviously the most appropriate voice for any AI agent !
I set it up with Piper and am using it in Home assistant. One problem I'm experiencing is that the last word is often truncated.
For example, if the sentence is "The garage lights have been turned off", I cannot hear the "ff" part, only the "o". Same with "on". This makes it impossible to differentiate the state, which is problematic in a smart home application. If I append the word "successfully", then the problem goes away. If I append the word "now", "now" itself is slightly truncated. I understand this is a common problem with this type of Piper model. But if there is any chance it could be fixed, I would greatly appreciate it.
In the meantime, I will try to work around this in the local LLM conversation agent I'm using by adding instructions in the system prompt.
First, I want to thank you for this amazing project. 2001 is my favorite movie of all time. HAL is obviously the most appropriate voice for any AI agent !
I set it up with Piper and am using it in Home assistant. One problem I'm experiencing is that the last word is often truncated.
For example, if the sentence is "The garage lights have been turned off", I cannot hear the "ff" part, only the "o". Same with "on". This makes it impossible to differentiate the state, which is problematic in a smart home application. If I append the word "successfully", then the problem goes away. If I append the word "now", "now" itself is slightly truncated. I understand this is a common problem with this type of Piper model. But if there is any chance it could be fixed, I would greatly appreciate it.
In the meantime, I will try to work around this in the local LLM conversation agent I'm using by adding instructions in the system prompt.