CoquiEngine is buffering small feed/last words in big feed texts #352
Description
Small part of the text(from the end) in feed just disappears. But those couple of words appear when I put something bigger next into the feed. Same goes for small phrases.
so, for example, when I put 10 words into the feed, it eats 2 last of them and plays them only in next big enough feed. and when I put just 3 words it eats all of them, playing them only if I will put bigger text into feed.
Reproduces with llm generators and with simple feed("text")
Tried many parameters, such as buffer_threshold_seconds, context_size and ect, but didnt worked(with defaults too). Still eating couple of last words in the sentence. Maybe I misconfigured something?