VAD Parameters
VAD controls you can tune
FlowbotAI exposes several tuning parameters under the Voice Activity Detection tab on the Agents configuration. These settings affect latency, interruption behavior, and sensitivity to quiet speakers or noisy environments.
Turn Endpoint Delay (ms)
This is the minimum time the agent waits before responding after the caller appears to stop speaking.
Precision note: Only multiples of 32ms are meaningful. Values between 1ms and 31ms produce the same behavior because the system evaluates audio in 32ms frames.
Trade-offs:
- Shorter delays → faster responses, but a higher chance of interrupting callers mid-thought.
- Longer delays → fewer interruptions, but the agent can feel slower or less responsive.
Minimum Turn Duration (ms)
This is the minimum duration of caller speech required to count as a valid turn.
Trade-offs:
- Shorter durations → captures very short caller audio.
- Longer durations → can ignore very short audio that might be noise, but may also ignore meaningful short replies like “yes” or “no”.
Minimum Interruption Duration (ms)
This is the minimum duration of caller speech required to interrupt the agent while it’s speaking. It’s similar to minimumTurnDuration, but is intended to be a stricter threshold so accidental noises don’t constantly interrupt the agent.
Important rule: If minimumInterruptionDuration is less than minimumTurnDuration, it is ignored.
Trade-offs:
- Shorter durations → more sensitive to interruptions, but may trigger on background noise.
- Longer durations → less sensitive to noise, but may miss legitimate interruption attempts.
Frame Activation Threshold
Valid range: 0.1 to 1.0. This is the threshold for treating an individual 32ms frame as containing speech.
Trade-offs:
- Lower thresholds → more sensitive to quiet speech, but increases false positives.
- Higher thresholds → less sensitive to noise, but can miss quiet or distant speakers.
Safe tuning practices
Start with defaults unless you have a measured problem
The default VAD settings work well for most applications. Adjust them only when you have specific, tested issues that can’t be solved through environment or hardware improvements.
The lowest-risk setting to modify
If you do tune VAD, turnEndpointDelay is typically the safest parameter to adjust. It is simply a practical starting value that works well in many scenarios—there is no “magic” number.
Make changes one at a time
- Change a single parameter → isolate what actually improved or regressed.
- Test in realistic conditions → include real users and the environments you expect in production.
- Watch the trade-offs → improvements in one area often introduce new issues elsewhere.
Try environmental fixes first
Before tuning VAD, consider improvements that reduce audio ambiguity:
- Audio quality → better microphones reduce the need for aggressive VAD tuning.
- Network quality → poor connections can create artifacts that affect perceived turn-taking.
Remember: VAD tuning is always a set of trade-offs. The system is designed to be intelligent and adaptive—trust the defaults unless you have a compelling, well-tested reason to change them.
Related Articles
Noise & VAD Overview
In this article we describe how FlowbotAI detects speech, suppresses noise, and keeps conversations fast and natural—without constant tuning. What this feature does in plain English Noise handling and Voice Activity Detection (VAD) are the “ears” of ...Explore Custom Tools
What “Custom Tools” mean in FlowbotAI Custom tools give your FlowbotAI agent the ability to communicate with systems outside the call—anything you can do in a function (or API endpoint) can be exposed to the agent as a tool. Tools are the foundation ...Explore Built-in Tools
What “built-in tools” mean in FlowbotAI FlowbotAI ships with a small set of built-in tools that cover common voice-application needs out of the box. They behave the same way as tools you create yourself: your agent can invoke them mid-conversation, ...FlowbotAI Tools Overview
What tools unlock in FlowbotAI Tools are the “action layer” for FlowbotAI voice agents. They let an agent do more than talk — it can look things up, update systems, and trigger workflows while keeping the conversation natural and on-track. In plain ...Embedding FlowbotAI in Netsapiens
This article explains how to enable FlowbotAI’s native Netsapiens integration inside the Netsapiens Office Manager Portal (OMP). Once enabled, Service Partners can activate, auto-provision, manage, and support FlowbotAI agents directly in their OMP — ...