Hi everyone, I’m using an ElevenLabs voice with my Vapi assistant. The audio sounds rushed, with almost no pause between sentences. I dug into the call logs and noticed that the transcriber is sending a lot of partials without punctuation, and even the final transcripts often cut off mid-sentence or don’t include a proper period. Could this be the reason ElevenLabs doesn’t respect sentence boundaries and just stitches everything together with micro-pauses? Has anyone else run into this, and how do you handle buffering between partial and final transcripts so the speech has natural pauses?

Hi Saltlife Kid, The issue you're experiencing with ElevenLabs likely stems from how Vapi is handling speech transcription and text normalization. The transcriber may not be properly placing punctuation in partial transcripts, causing the text-to-speech engine to treat the continuous stream of text as a single run-on sentence. This could explain the lack of natural pauses. 1. Text Normalization Processors: Consider configuring the

formattersEnabled

array in your

Assistant.voice.chunkPlan.formatPlan

to control text normalization. This can help ensure that punctuation and other sentence boundaries are better respected by ElevenLabs. Learn more about this feature in the [February 20, 2025 Vapi changelog](https://docs.vapi.ai/changelog/2025/2/20). 2. Latency Considerations: Vapi's Neuphonic voice synthesis feature mentions additional processing latency, which might not be directly related to your issue but is worth testing if changes in latency impact the pacing. 3. Review Transcription Settings: Check if your transcription settings, such as

disablePartialTranscriptions

or confidence thresholds, are optimized to provide complete and punctuated transcriptions. This might help improve punctation and sentence boundary detection. 4. Word Boost and End Utterance Silence: Utilize the

wordBoost

feature and adjust

endUtteranceSilenceThreshold

for better handling of speech patterns and pauses. Implementing these settings may help enhance the quality of your assistants' audio responses and resolve the rushed speech issue.