Radim Kopp
03/28/2025, 11:16 AMVapi Support Bot
03/28/2025, 11:17 AMsilenceTimeoutSeconds
- responseDelaySeconds
- llmRequestDelaySeconds
3. Endpointing Settings:
- Try setting Deepgram's endpointing to 300ms
- Set startSpeakingPlan.waitSeconds
to 0.8 seconds
- For non-English transcriptions, set smartEndpointingEnabled
to false
4. Voice Model Settings:
- Try adjusting these voice model parameters:
stability: 0.5
similarityBoost: 0.75
- Consider switching to a different voice provider for more stability
- The "flash model" has known hiccups - try using the "turbo model" instead
5. Resource Utilization:
- Monitor your server's CPU, memory, and bandwidth usage
- High call volumes can sometimes slow down response timesVapi Support Bot
03/28/2025, 11:18 AMRadim Kopp
03/28/2025, 11:26 AMShubham Bajaj
03/28/2025, 4:44 PMRadim Kopp
03/28/2025, 4:50 PMKyle Brunker
03/30/2025, 10:16 AMRadim Kopp
03/31/2025, 8:56 PM