On the road

Setups that work, what to expect from each, and where it stops working.

Before the first ride

Do this at home, not in a car park with gloves on.

  1. Pair the headset and check the phone is routing to it — play something and hear it in the helmet.
  2. Add the server by hand, by scanning a QR code from whoever runs it, or by following a mumble:// link.
  3. Set the profile to Helmet and the mode to Voice activated.
  4. Set the gain with Test microphone, wearing headphones, with the helmet on. Peak around three quarters when you speak normally.
  5. Connect, and leave it connected while you get ready. The app takes the audio session only while a call is up, so nothing is recorded and your headset keeps its sound quality for other apps until you connect.

Two riders

The simplest case, and the one an intercom already handles well. Use this instead when you keep losing each other — a city, heavy traffic, or a ride where one of you stops for fuel and the other does not.

Expect: a couple of hundred milliseconds of delay. You will talk over each other for the first few minutes and then stop doing it. Neither of you needs to be within any particular distance of the other.

A group

Where it earns its place. Everyone joins one channel, and the group can be strung out over several kilometres without anyone dropping out of the conversation.

A group strung out along a road, all on one channel Five riders spread over eight kilometres of road. A Bluetooth intercom would connect only the two who happen to be close together, breaking the group into fragments. Every rider instead holds their own connection to the same Mumble server, so the spacing between them carries no meaning at all. FIVE RIDERS, EIGHT KILOMETRES one channel, one server back front stopped for fuel what an intercom would still reach
This is the case the app was built for. A group does not ride in formation: somebody stops for fuel, somebody catches a red light, somebody is simply faster. On an intercom the group silently breaks into fragments that can each hear themselves. On one channel, the spacing means nothing.

Rider and passenger

Both wear headsets and both join the channel. It works, with one caveat: two microphones inside two helmets a foot apart, both on the same channel, is the easiest way to build a feedback loop.

How a rider and passenger build a feedback loop Two helmets a foot apart on the same channel. The passenger's speaker is heard by the rider's microphone, which sends it back to the passenger's speaker, which the rider's microphone hears again — a loop that climbs until it howls. Turning one speaker down breaks it. RIDER AND PASSENGER, ONE CHANNEL rider passenger speaker into the other's microphone and straight back again BREAK IT ANYWHERE 1 Turn one speaker volume down. 2 Feedback suppression: cut when a howl builds. 3 Failing that, put the passenger on push to talk. A loop needs every step. Removing one is enough.
Two microphones inside two helmets a foot apart, both on the same channel, is the easiest way there is to build a feedback loop — each speaker reaches the other's microphone and the gain goes round. It only takes one step of that circle to break it.

If you hear a howl:

  1. Turn one of the two speaker volumes down.
  2. Set Feedback suppression to Cut only when a howl builds on both.
  3. Failing that, put the passenger on push-to-talk.

Riding with music

Music and voice activation are an unhappy pair today: sharp, tonal, plucked notes open the gate, and the far end hears your music instead of you.

Until that is fixed, either:

Music from your phone ducks under a call rather than stopping — that part works — but ducked music is still music, and the detector still hears it.

Being heard over your own music is the half that now works. The gate takes its threshold from the noise floor it has measured rather than from a fixed level, so music raises the bar instead of drowning the speech that has to clear it: on a recording of talking over music, the helmet profile went from letting 63% of the speech through to 98%. Speaking over music is no longer the problem. Music speaking for you still is.

A rally, or riders who do not know each other

Where it stops working

No signal, no conversation. A tunnel, a valley, a mountain pass, a border where your data plan quietly stops. The app reconnects on its own when the link comes back, and the elastic buffer catches you up rather than leaving everybody a second behind — but while there is no data there is no voice, and an intercom would still have been working.

What happens when the signal goes A road passing through a tunnel, in three stages. Before it there is coverage and the conversation works. Inside there is no data and no voice at all — the honest limitation. On the far side the app reconnects on its own and plays the backlog off faster than real time rather than leaving everybody a second behind. NO SIGNAL, NO CONVERSATION coverage no data back on the air the conversation works nothing gets through reconnects, and catches up 2×
This is the defining weakness and no setting fixes it. What the app does do is come back on its own and catch up rather than leaving the whole group permanently a second behind — but while there is no data there is no voice, and that is exactly where an intercom would still be working.

Two practical consequences:

Data and battery

Roughly 3–6 MB per hour of actual talking, more when the link is poor and error correction rises. Silence costs almost nothing — voice activation means nothing is sent between sentences.

Battery is the more noticeable cost. A phone holding a Bluetooth call and a mobile data connection for several hours will want charging. A hardwired mount solves it and is worth having anyway.