AI Engine Impact
Your choice of AI engine affects perceived call quality:Dualplex
- Lowest latency of the three engines
- Quick back-and-forth
- Best for most calls
Pipeline, with filler audio
- Speech is transcribed, the model answers, the reply is spoken, so there is a pause to cover
- Filler audio fills that pause with short phrases, which keeps the call feeling natural
- Filler audio only exists on this engine
Speech-to-speech
- Native voice generation
- No filler audio, because there is no gap to fill
Optimization Tips
- Start with the Dualplex engine
- If you stay on Pipeline, turn on filler audio to cover the pause
- Monitor call recordings for quality
- Test different voice models
Quick Fixes
If experiencing issues:- Save the assistant, then try a call from Test assistant — tests run on the saved version
- Switch to a faster engine setting
- Enable filler audio
- Check network connection
Tip: Dualplex gives the best perceived call quality on most assistants. If you stay on Pipeline for a particular voice, turn filler audio on.