Compare original and processed singing, dialogue and music examples. Listen to what remains before deciding what works for your recording.
The music-removal examples below are project-owned synthetic editorial tests—not customer projects, testimonials or a promise about every source. The voice examples reuse two parent mixes; they are not five independent recordings or a controlled benchmark. The October 15 tests used the music-removal pipeline available at their review date, with no fabricated processed stems. Both players use MP3 delivery encoding. Delivery durations include codec padding; source/result review dates below refer to the existing approval, not a new listening review. Your own file's preview is the better purchase decision.
REPLACE MUSIC · CONTROLLED EXAMPLE
From an old music bed to a new mix
Delivery: 13.20s MP3Processing: Separation, then local music remixSource/result review: UTC
This 13-second test combines recorded audiobook narration and instrumental music. Compare the separated voice before hearing the new music: new music can mask old-music residue without removing it.
Sample limit: This is a controlled mixture, not a customer project or an unedited real-world scene. It does not demonstrate sound-effects preservation, overlapping singing or spontaneous conversation. Preview your own difficult section before purchasing.
Speech: LibriSpeech test-clean recording 6930-75918-0005, from the corpus prepared by Vassil Panayotov with Daniel Povey, derived from LibriVox audiobooks. LibriSpeech / OpenSLR SLR12, CC BY 4.0.
Changes: excerpts combined and level-adjusted, AAC source encoding, music separation, local mixing with a different instrumental track and fades, and MP3 delivery encoding. No endorsement by the source contributors is implied.
HEAR WHAT CHANGES
Less voice. Listen to what remains.
Two 12-second controlled mixes from licensed recordings. These are editorial tests, not customer projects or a promise about your video.
REMOVE SINGING
Singing over music and sound effects
Listen for the singing, then check the musical bed and three added tone effects.
Limit: This separates a vocal stem; it can also reduce spoken voices. Reverb or other sounds may change.
Sources, licenses and processing changes
A chantar m’er — Comtessa de Dia; performed and recorded by Makemi (2006), CC BY 2.5. Music: Cipher — Kevin MacLeod (incompetech.com), CC BY 4.0. Changes: excerpts, level adjustment, controlled mixture with generated tone effects, AAC source encoding, vocal separation and MP3 delivery. No endorsement implied.
Actual model output with MP3 delivery encoding; no extra cleanup or loudness normalization. Controlled test reviewed November 3, 2026. Not a natural scene or a customer endorsement.
REMOVE DIALOGUE
Spoken narration over music
Listen for the narrator, then check the musical bed for changes and watery artifacts.
Limit: Speech can remain when voices overlap singing or music. Separation can also change sounds you want to keep.
Sources, licenses and processing changes
Speech: LibriSpeech, prepared by Vassil Panayotov with Daniel Povey from LibriVox audiobooks; recording 5639-40744-0032 (OpenSLR SLR12), CC BY 4.0. Music: Cipher — Kevin MacLeod (incompetech.com), CC BY 4.0. Changes: excerpts, level adjustment, controlled mixture, AAC source encoding, dialogue separation and MP3 delivery. No endorsement implied.
Actual model output with MP3 delivery encoding; no extra cleanup or loudness normalization. Controlled test reviewed November 3, 2026. Not a natural scene or a customer endorsement.
Delivery: 15.02s MP3Processing: Music removal; retained non-music residualSource/result review: UTC
Compare the generated impacts and room-noise texture before and after music reduction. Listen to the impact tails, not only the quieter score.
Project-generated score, noise and impact sounds; no speech or customer recording.
Sample limit: Quiet ambience and transient texture can change or disappear. This procedural test does not validate natural footsteps, applause, crowds or every important effect.
Delivery: 15.02s MP3Processing: Music removal; retained non-music residualSource/result review: UTC
Compare quiet syllables and changes in voice texture in this deliberately difficult synthetic dialogue-and-music overlap. Reduced music is not a guarantee of clean original speech.
Existing project-owned synthetic dialogue mix with added procedural music and impacts.
Sample limit: This shares a mixed parent with the earlier hard-mix example. It is not a field interview, independent speaker sample or proof that overlapping voices will survive.
Delivery: 15.02s MP3Processing: 48 kbps source round-trip, then music removalSource/result review: UTC
The source passed through a 48 kbps MP3 encode/decode before separation. Compare consonants, breaths and pauses; separation cannot restore detail already lost to compression.
Existing project-owned synthetic voice-and-music mix, procedural additions and a low-bitrate round-trip.
Sample limit: This shares a parent voice with the earlier easy-mix example. It does not validate every codec, singing voice or long podcast. Both players use delivery MP3 encoding.
Delivery: 13.04s MP3Processing: Earlier music-removal resultSource/result review: UTC
Compare the same phrase in both players. Listen to the voice’s consonants and pauses, then ask whether the quieter music has been reduced without making the voice sound hollow.
Delivery: 12.04s MP3Processing: Earlier music-removal resultSource/result review: UTC
A useful result may still contain musical residue or changes in voice texture. Listen beyond the loudest word: check whether quiet syllables and scene details survive. Reduced music is not a guarantee of clean original dialogue.
Use the original as a reference, not just the processed result in isolation. Compare the same moment at similar dialogue loudness and with the same headphones or speakers. A louder processed file can seem clearer without actually preserving more detail.
Words: can you understand names, numbers, consonants and quiet endings without guessing?
Voice texture: does the speaker sound stable, or does the result become metallic, watery or hollow?
Context: do room tone, movement, applause and essential effects still belong to the scene?
Music residue: is what remains acceptable for the next edit, including pauses between words?
Listening fatigue: does the result remain comfortable when played continuously, rather than in a quick toggle?
A free preview covers up to 15 seconds and no more than half your file. Minute-balance previews cover up to 30 seconds and use 30 seconds of balance; existing purchases may retain included previews. Move the preview selector to the difficult overlap rather than judging a long recording by a clean intro. Read the preview decision guide for a fuller checklist.
From acceptable preview to a full export
For a video file, music reduction and video packaging are separate checks. The tool processes extracted audio and combines the result with the local video. Container extension alone does not prove codec support. Keep the original available on this device and let the browser check compatibility before checkout.
Before paying, review the pack price and the file’s actual-duration requirement. A four-minute-and-one-second file uses four minutes and one second; replaying the preview does not mean buying a full export. After processing, check synchronization at both the start and the end, inspect important transitions, and actually download and play the exported file in its intended destination.
A result marked ready is not a permanent backup or proof of a saved download. Results are time-limited; download promptly. See the MP4 audio-replacement and export guide for the difference between preserving video packets and changing the soundtrack.
Not every test passes. We held back another synthetic clip because important effects were lost even though speech remained. Audible dialogue alone is not a successful result when scene sounds matter.
If the preview loses important words, removes essential effects, or introduces more distracting artifacts than the music, stop. Use an original project, stems or music-free master when available. A different mix may require manual dialogue editing rather than stronger separation. Our examples do not establish suitability for your recording.
Found a confusing result or export problem? Use Feedback or contact support without attaching media. We can use your description to improve the workflow, but do not publish it as a customer endorsement without permission.
Example facts updated . Removal tests reviewed October 15, 2026 · Licensed replacement example reviewed October 24, 2026. Vocal examples reviewed November 3, 2026. Testing and editorial policy.
Music and Vocal Removal Examples | RemoveBackgroundMusic
RemoveBackgroundMusic