AI Voice Isolator - Separate Voice from Music
AI voice isolator: strip background music and noise and keep clean vocals. Plus a voice changer and forced alignment for exact subtitle timing.
最終更新: 2026-10-18
The VoiceSmiths AI voice isolator separates the voice in a recording from its background music and noise, and hands you back clean vocals. It sits next to two more tools for audio you already have: a voice changer that keeps the performance and swaps the voice, and forced alignment that turns a recording plus its script into exact subtitle timing. All three live in the Audio Tools tab of the studio and take files up to 200MB.
At a glance
| Tool | Input | Output | Typical use |
|---|---|---|---|
| Voice isolator | A recording | The voice, with background music and noise removed | Clean cloning samples, rescue field audio, swap the music |
| Voice changer | A recording + a target voice | The same performance in the target voice | Act it yourself, then swap in a character voice |
| Forced alignment | A recording + its script | Start and end time of every word, SRT export | You have the script, just not the timing |
Every tool shows its credit price on the button, and failed jobs refund automatically.
AI voice isolator: pull clean vocals out of the mix
Strip background noise and music, keep the voice. Frequent uses:
- Clean cloning samples: noisy recording? Isolate first, then clone — quality improves noticeably
- Rescue field audio: street interviews and event recordings become usable again
- Swap the soundtrack: pull the voice off a video whose music doesn't work, then lay in a new score from the AI sound effects generator
- Prep for transcription or alignment: when the background is loud, clean the voice first, then run speech to text or forced alignment
How to use it
- Open Voice Studio → Audio Tools
- Upload the recording (up to 200MB)
- Click Isolate; the button shows the credit price before you spend anything
- Listen to the result and download the clean vocals
What it won't do
The output is the isolated vocal, not the backing track with the voice taken out. If you need a karaoke instrumental, this AI voice isolator isn't the tool for it; if you need a clean voice you can actually use, it is.
Voice changer: keep the performance, swap the voice
Upload a recording, pick a target voice — pacing, pauses and emotion survive intact; only the timbre changes. Great for: performing a line yourself to control the delivery, then swapping in the character's voice; or fanning one dry read out into several character versions.
Real before/after · the same read, before (Sarah) → after (Callum):
Before:
After (identical performance, different voice):
Music under the take? Isolate the vocals first, then run the clean voice through the changer.
Forced alignment: recording + script = exact timing
Already have both the audio and the script, just missing timestamps? Forced alignment returns each word's precise start and end. Real output from our English homepage sample:
The [0.08s→0.14s] rain [0.16s→0.34s] had [0.36s→0.50s]
stopped, [0.52s→0.80s] ...
Versus speech to text: transcription guesses what was said; alignment already knows, and marks when — faster, more precise, zero typos when you have the script.
Where to find the AI voice isolator and the other tools
All three live in Voice Studio → Audio Tools: upload a file (≤200MB), the credit price sits on the button, failures refund automatically. Alignment results download as SRT.
FAQ
Can the AI voice isolator take out background music? Yes. It removes background music and noise together and leaves the clean voice.
Do I also get the instrumental track? No. The output is the isolated vocal, not the backing track with the voice taken out, so it is not a karaoke maker.
Is there a file size limit? Up to 200MB per file, for all three tools.
How is it priced? In credits. Each job shows its credit price on the button before you start, and failed jobs are refunded automatically. See pricing.
How is forced alignment different from transcription? Transcription guesses what was said; alignment already knows and marks when it was said. When you have the script, alignment is faster, more precise and has no typos.
Clean up your first recording
Got a voice buried under music or noise? Run it through the AI voice isolator and get the clean vocals back in a few steps.