MulingStream is built for the spoken word: the sermon, the welcome, the prayers, the announcements. Singing is something else. Here is what really happens when the music starts, what works, what does not, and the one choice at the sound desk that makes the biggest difference.
Sooner or later every church that uses live translation asks it: and what about the songs?
It is a good question, because worship is not a small part of a service. In many churches it is a third of the morning or more. A guest who follows the sermon in their own language would love to follow the songs too. So we want to answer it carefully, saying what works, what does not, and what we recommend.
MulingStream translates the spoken word. It is built for a person talking: the sermon, the welcome, the prayers, the Bible reading, the announcements. It does not translate songs as songs. It does not sing, it does not make lines rhyme, and it does not fit the words to the melody.
What happens during the singing depends almost entirely on one thing: what MulingStream can hear.
This is the setup we recommend, and with it the songs mostly stay out.
When MulingStream is fed from the speaker's or the pastor's microphone, or from a channel at the sound desk that carries only the speaking microphones, the band and the congregation barely reach it. During the worship the listener's screen simply goes quiet. When the worship leader speaks between songs, prays, or reads a verse into a microphone that is in that feed, those words come through as normal. When the pastor walks up to preach, the translation is already running, and nobody had to touch anything.
At the sound desk this is usually an aux send, a spare bus or a matrix output with only the speech microphones in it. The main mix works too, but it brings the music with it.
If the input is the main mix with the band in it, or a microphone in the middle of the room, MulingStream will try to catch words from the singing, because it is listening for words. Here is what that looks like:
Sung words are caught less reliably than speech. Over drums, guitars and a full congregation, many lines are missed or caught in pieces. Clearly sung lines over a quiet accompaniment come through better. What is caught is translated as plain text. A line of a song arrives on the guest's screen as a translated sentence, not as a lyric that scans or rhymes. The voice reads it, it does not sing it. A guest listening with an earphone hears the translated line spoken, which can feel odd in the middle of a song. Guests who find it distracting can stop the voice and simply read, or stand and sing with the room. Repeats are kept down. In a church room, short repeated lines such as "amen" or "hallelujah" said or sung again and again are not sent to the guest over and over, and small fragments are joined to what follows so they make sense. We added this in July for prayer and worship, where people speak in short phrases with long pauses between them. A line from Scripture can bring up the verse. Many worship songs quote the Psalms or the Gospels. When a sung line clearly quotes a verse, Smart Hints may show that verse in the guest's own language, from a Bible translation in that language.
So it does not break, and nothing bad happens to the service. But it is not a good way to give a guest the songs, and we would rather a church knew that beforehand.
MulingStream is built for the spoken word. For the songs, the guest is part of the room like everyone else.
We recommend starting the translation before the service and letting it run to the end, songs included.
With a church pass, one pass covers the whole gathering, so leaving it on through the worship costs nothing extra. A church room waits a full hour without speech before it stops by itself, so a long time of worship does not end the session. If the sound desk does need to stop it, starting again is one click, and the guests stay in the room.
If your guests should be able to follow the words of the songs, the most reliable way today is the one many churches already use: show the lyrics on the screen. A guest who does not know the language can still sing along by reading the words, and for many songs they will recognise the tune. If you show translated captions on a screen with MulingDisplay, it can run on a second screen next to the song slides, so the projector keeps doing what it always did.
And many guests tell us the worship is the part they need translation for least. They know what is happening. They feel the room. It is the sermon they want to understand, and that is what MulingStream is for.
Works well: everything spoken, including between songs, prayers and readings. Works partly: sung words, if the music reaches the microphone. Lines come through as plain translated text when they are caught. Does not do: sing, rhyme, or translate a song as a song. Our recommendation: feed it the speaking microphones, leave it running, and show the lyrics on the screen.
More on the sound side is on sound and media and for the person at the desk on how it works for leaders.
From 3 October 2026 Smart Hints treat a song as one thing. A sung line now brings up only its Bible source, if it has one, and only once per song, instead of a note for every line that echoes Scripture. Words inside a song are understood as the song, not as slang or a separate saying to be explained.
From 20 October 2026 the speaker's screen has a Mute switch under the record button. During the worship, the person at the desk can mute the microphone instead of stopping and starting again. Listeners see a small "Paused" note so they know nothing is wrong, and when the microphone is unmuted the translation carries on at once. On a church pass nothing changes about what is covered.
Every church's sound is different. We show you the whole thing live, in about fifteen minutes, and look at your setup with you. Nothing is decided on the call and nothing follows it: you see it, and then you are the judge.