Wedding speeches and vows · Updated 1 October 2026
Less music.
More of the speech you want to keep.
If a DJ’s music is louder than a wedding toast, turning down the whole clip turns down the speaker too. Voice separation gives you a way to try reducing the music while keeping the speech.
Use your own audio or video file. Listen to the result before deciding whether it is worth processing the full speech.
Start with the best recording you have
Music added in the wedding edit: ask the videographer for a version without the music, or lower its separate track in the editing project. This preserves the recorded voice without asking AI to separate it.
Music played in the room: it is mixed into the microphone recording. Compare the camera audio, a close microphone or recorder, and any recording from the sound desk. Start with the version where the words are clearest.
Only a phone clip: use the original file if available. Keep an untouched copy while you test the result.
Hear a voice-and-music example
This is film dialogue with music deliberately added, not a wedding recording. It demonstrates the same separation task. The after is actual Mirinex output, and both samples use the same gain.
Your room, microphone and music will sound different. Use your own-file result to judge it. Demo credits.
Separate the speech, then choose what to keep
- Choose the speech section. Each file can be up to 20 minutes and 250 MiB. If you have a full reception recording, export the toast from your video editor first. Keep a little room sound before and after the speech so you can hear the change at the edges.
- Open Voice isolator and choose the file. Supported audio and video include MP4, MOV, MP3, WAV and M4A, depending on browser codec support.
- Compare Voice with Original. Listen to names, quiet words, the loudest DJ passage and any moment where people speak at once.
- Bring back some of the room if it helps. Raise the Music fader slightly and compare again. The removed track can contain applause and ambience as well as music. A little background may sound more natural than an isolated voice.
- Download the mix. For supported video, the output combines your picture with the adjusted audio. If the browser cannot export that video format, download the WAV fallback and place it under the original picture in your editor.
Music, other voices and distortion need different expectations
A singer or another guest talking: voice isolation may keep them too. It does not select the bride, groom or a particular speaker from overlapping voices.
Wind, hiss or electrical noise: compare the separate noise-removal tool. Do not keep processing a voice that is already clear; each pass can change it.
A distorted microphone or missing words: lowering music cannot promise to repair clipping or recover words that the recording did not capture clearly. A cleaner alternate microphone recording is often the better starting point.
For an important memory, listen through the downloaded file with the original nearby. Check the laugh, pause or applause you wanted to preserve—not just the first sentence.
Try it on your speech
Choose the recording and hear the speech come forward in moments. Keep as much of the room as you like, then download the mix.
Hear your speech with less music
Hear the first minute of your own file. A Day Pass or Pro processes the whole thing.
Related recording help
For a video you plan to upload, see removing music from a YouTube video. For a noisy spoken recording, see podcast and interview cleanup.
Demo credits and changes
Film audio: “Tears of Steel,” © Blender Foundation, mango.blender.org, CC BY 3.0. Excerpt from 00:25–00:41.
Added music: “I Want To Destroy Something Beautiful” by Josh Woodward, track source, CC BY 3.0 US. Excerpt mixed under the dialogue.
The constructed mix was processed with AI voice separation. Both samples share the same −0.6502 dB gain and identical silent padding to 16 seconds. No endorsement by the filmmakers, performers or musician is implied.