PonToAmbient User Guide

Background

Why I built PonToAmbient

For a while, music had become something I simply left playing in the background. Wanting to rediscover it as something to listen to attentively, I returned to dedicated listening through a DAC and headphones. The added immersion was wonderful—but it also meant missing the everyday sounds that matter, such as family calling or a pet crying out.

I wanted to preserve the fidelity and immersion while allowing only essential sounds from daily life to break through with minimal interruption. PonToAmbient was created from that idea.

Quick Start

Up and running in five steps

Complete the safety check with music paused, then test at your normal listening volume.

Select input and outputChoose the microphone and the USB DAC or headphones you want to use. The selection also becomes the macOS system-wide input and output.
Use Standard microphone modeVoice Isolation removes surrounding noises. Select Standard in Control Center if you also want loud-sound detection.
Start MonitoringAllow microphone access when macOS asks. Configure devices while stopped and fine-tune sliders live while monitoring.
Run the three-second ambient testBegin with a low ambient level. Be especially careful about feedback when using speakers.
Set music volume in the music appBalance the music and passed-through ambient sound independently.

PonToAmbient stopped screen showing input, output, voice sensitivity, loud-sound threshold, and ambient output controls. — Stopped. Select devices and run calibration in this state. Click the image to open it at full size.

Screen Status

Monitoring and output are different states

The microphone is not sent to your headphones just because monitoring is active.

Monitoring · Output Off

The microphone is being analyzed, but ambient sound is muted while the app waits for a voice or a sound above the threshold.

Ambient Output Active

A voice or loud sound opened the path from the microphone to the selected output. The trigger reason appears on screen.

Blue: average input level (RMS) White: instantaneous peak Orange: loud-sound threshold

The Silero percentage is the AI’s current voice probability. At Standard sensitivity, 65% is the detection threshold. Voice detection is independent of the loud-sound threshold, so a quiet call can still open the ambient path.

PonToAmbient monitoring screen with live input level and Silero voice probability. — Monitoring. Slider changes take effect immediately, so the separate Calibration button is not shown. Click to enlarge.

Real-world Example

The settings used by the author

A practical starting point from a setup using a WALKMAN as a USB DAC.

Tsuyopon’s example

Amazon Music app volume is set to 30, with the PonToAmbient settings shown in the screenshots.

Input: Studio Display XDR mic
Output: WALKMAN / 192 kHz
Detect Human Voice: On
Voice Sensitivity: Standard
Low Latency During Conversation: On
Detect Loud Sounds: On
Loud-sound Threshold: -19 dBFS
Ambient Output Volume: +12 dB
Maximum Ambient Level: -12 dBFS
Hold Output: 3 seconds

These values match the author’s microphone position, WALKMAN, headphones, and listening level. Do not begin at +12 dB—start low and raise the ambient level carefully.

Controls

What each setting does

Think of detection and listening level as two separate groups of controls.

Detect Human Voice

Silero AI opens the ambient path whenever it recognizes human speech, regardless of the loud-sound threshold.

Voice Sensitivity

High reacts most readily. Approximate Silero thresholds are Low 85%, Standard 65%, and High 45%.

Low Latency During Conversation

The first call uses the 400 ms pre-roll. Repeated speech switches to low latency at a safe silent point.

Detect Loud Sounds

Passes any sound over the threshold, whether or not it is speech—useful for doorbells, claps, and household sounds.

Loud-sound Threshold

Move left to react to quieter sounds or right to require louder sounds. Changes are immediate while monitoring.

Ambient Output Volume

Adjusts microphone audibility from -24 to +12 dB. It does not change the music app’s volume.

Maximum Ambient Level

Preserves gain for quiet speech while limiting strong peaks such as sneezes. Lower values apply a stronger ceiling.

Hold Output

How long the ambient path stays open after the latest detection. Short values return to music sooner.

Launch at Login

Keeps the app available from the menu bar after login. Monitoring still starts manually for safety.

Calibration

Make the sound you want to catch

Calibration is available while stopped. Voice analysis pauses so you can focus on the loud-sound threshold.

Click CalibrationPause music and leave the Mac and microphone in their normal positions.
Make the target soundRing the doorbell, clap, or type from the real listening distance.
Set the boundaryUse the Catch / Ignore result and maximum peak reading to keep only the sounds you need.
Finish calibrationStart monitoring. You can still fine-tune the threshold live with the slider.

Voice detection stabilizes during the first minute

Silero voice detection starts immediately, and a clear human voice can open the output without waiting. During roughly the first minute, it also adapts to steady room sound to suppress startup false positives. Loud-sound threshold detection is independent of this stabilization and follows your selected threshold from the moment monitoring starts.

dBFS is not room sound pressure

It is the digital level reaching your Mac’s microphone. The reading changes with microphone type, distance, direction, and mic mode.

Conversation Mode

Catch the first word, then talk naturally

The app combines a 400 ms pre-roll with an automatic low-latency conversation path.

The first call

The app continuously holds about 400 ms of delayed microphone audio. When AI detects speech, it can open the path without cutting off the beginning.

Ongoing conversation

Repeated voice detections switch to low latency at a safe silent point. The 400 ms path returns after the conversation ends.

Loud sounds alone do not start conversation mode

This prevents typing or repeated claps from switching the audio path to low latency.

Listening Safety

Raise quiet speech, constrain peaks

Use Ambient Output Volume and Maximum Ambient Level together.

Output volume = audibility

Controls how much quiet voices are amplified. Start low and raise it until normal calls are easy to hear.

Maximum level = ambient ceiling

Restrains strong peaks such as sneezes. It affects only PonToAmbient’s microphone path—not Amazon Music or other system audio.

This is not a guaranteed ear-level limit

The dBFS ceiling is a digital signal limit. Actual listening level still depends on your WALKMAN, DAC, amplifier, and headphones. Test with music paused and hardware volume low.

Troubleshooting

When something is not working

The input meter does not move

Allow PonToAmbient in System Settings › Privacy & Security › Microphone. While stopped, select the input again. Stopping and restarting monitoring can also restore the route.

Human voices are not detected

Select Standard microphone mode and watch whether the Silero percentage rises with speech. If distant speech is missed, choose High sensitivity and restart monitoring.

I hear typing below the threshold

After a voice or loud sound opens the path, all ambient sound remains audible for the Hold Output duration. If the path opens incorrectly, move the loud threshold right or lower voice sensitivity.

Loud-sound detection reacts too much—or not at all

Run calibration while stopped and make the real target sound. Move the slider left to react more easily or right to require a louder sound.

Monitoring stopped after I changed output

The app mutes while following a new route. Switching to built-in speakers causes a safety stop to prevent feedback; verify the output and restart monitoring manually.

Music and ambient sound will not play together

Your music app may be using exclusive output. Select a mode that allows the output device to be shared with other apps.

How do I reopen the window?

Choose Show PonToAmbient from the menu-bar icon. Closing the window does not quit the resident app.

Privacy

Your microphone audio stays on your Mac

On-device analysis only

PonToAmbient uses its bundled Silero model for voice detection. It does not record, save, or upload microphone audio, and it does not use Apple Intelligence or cloud AI.