Most (if not all) of the unnatural pauses in the TTS generated output are a direct result of the LLM using commas all over the place. It might be proper grammar, but we need to add this to the system prompt to NEVER use commas, unless it is needing a pause, but even then, we should be using [pause:0.6s] explicitly instead of relying a side effect of using a comma.
Most (if not all) of the unnatural pauses in the TTS generated output are a direct result of the LLM using commas all over the place. It might be proper grammar, but we need to add this to the system prompt to NEVER use commas, unless it is needing a pause, but even then, we should be using [pause:0.6s] explicitly instead of relying a side effect of using a comma.