Sound Servo Lag And Mouth Twitch On ESP32

Yes - what you’re describing is usually not WiFi latency from the ESP32, since the servo is snappy when driven directly. It’s typically a mix of the Sound servo skill’s update timing, the audio capture/sampling behavior, and a little bit of low-level noise near silence.

What to check first

1) Sound servo has an update rate

Open the Sound servo (PC Speaker) skill’s settings from the kebab menu and look at the Update Speed value.

• Lower number = more responsive
• Default is usually 100 ms
• If yours is set higher, that can absolutely create a noticeable lag

Try setting it to 100 ms or even slightly lower if the skill allows it.

Skill docs:

• Sound servo (PC Speaker)

2) servo speed can add lag

Yes, servo speed can cause the jaw to "follow" slowly and make the mouth feel delayed.

For a lip-sync mouth, I recommend explicitly setting the servo speed to the fastest response at startup:

Servo.setSpeed(D14, 0);

You can put that in an Initialization Script or a startup script so it always applies when the project loads.

If you want to verify the current setting:

print(Servo.getSpeed(D14));

Fixing the tiny twitches near silence

That twitching is usually because the audio level never truly reaches zero - the sound servo is reacting to tiny fluctuations in the audio floor.

Best practice: add a noise floor / deadband

The cleanest approach is to not rely on the Sound servo skill alone if you need tighter control. Instead:

1. Use the skill to detect audio level
2. Read the level from its variable
3. Apply your own deadband / threshold
4. Drive the servo only when the value is meaningfully above silence

If you want, you can also build this as a small custom robot skill in ARC using JavaScript or Python:

• Create a Robot Skill

Simple deadband script example

If the Sound servo skill is writing its detected level to a global variable, you can use a script to "hold still" below a threshold.

Example JavaScript logic:

var level = getVar("$AudioLevel", 0);
var deadband = 8;

if (level < deadband) {
  Servo.setPosition(D14, 90); // replace 90 with your closed-mouth position
} else {
  var mapped = Utility.map(level, deadband, 100, 90, 130);
  Servo.setPosition(D14, mapped);
}

Notes:

• Replace $AudioLevel with the actual variable name used by your Sound servo skill
• Replace 90 / 130 with your real closed/open mouth positions
• If the skill output range is different, adjust the mapping accordingly

Even better: smooth the value a little

If the jaw still jitters, add a small smoothing filter so tiny spikes don’t move the servo.

var level = getVar("$AudioLevel", 0);
var deadband = 8;
var lastPos = getVar("$JawPos", 90);

if (level < deadband) {
  level = 0;
}

var target = (level == 0)
  ? 90
  : Utility.map(level, deadband, 100, 90, 130);

var newPos = (lastPos * 3 + target) / 4;
setVar("$JawPos", newPos);
Servo.setPosition(D14, newPos);

Why the lag may still exist

If the delay is truly 1-2 seconds, that’s longer than I’d expect from the skill alone. A few things to check:

• Update Speed in the Sound servo skill
• Whether the PC is outputting audio to a device that ARC is actually monitoring
• Whether the servo is getting commanded by another skill or script at the same time
• Whether the skill’s min/max and scalar are causing it to "ramp" in a way that feels delayed

Also, note that Sound servo (PC Speaker) reacts to your PC’s system audio output, not the ESP32’s audio. If your TTS/music is routed through a device with any processing or enhancement, that can change the feel.

My recommendation for the smoothest mouth on ESP32

1. Set servo speed to 0

   Servo.setSpeed(D14, 0);
   

2. Set Sound servo update speed to 100 ms

3. Use a deadband

4. Optionally smooth the motion with a small filter

5. Make sure only one thing is controlling the jaw servo

Best ARC approach

If you want a really polished mouth with less twitch and better control, I’d actually recommend moving from the stock Sound servo skill to a small custom JavaScript solution that reads the audio level and drives the servo with your own threshold/smoothing rules. That gives you much more control than the built-in behavior.

If you want, I can write you a ready-to-paste ARC JavaScript script for:

• a noise gate
• a smoothed jaw position
• and a closed-mouth fallback when audio is silent

If you share the exact variable name the Sound servo skill is using, I’ll tailor it for your project.