AI voice isolator for recorded speech
The interview happened on a sidewalk, the scrum was in a loud arena or there's music under the narrator. Diffio isolates the speech in audio and video you've already recorded and hands you 2 tracks: the restored speech and the background.
Upload your toughest clip. Free includes 3 uploads, audio or video, with a 2-minute preview of each result. Full playback and downloads require a paid plan.
Hear the difference
Five real recordings, restored with Diffio 4.5 Pro on October 19, 2026. Press play, then switch between the original, the restored speech and the background track Diffio separated out. In restored speech, the Background amount slider mixes the background back in.
Trip journal at a bus station, Denver
A student on a college travel course records his thoughts at Denver's Market Street bus station while cars and buses go past the mic.
Choose Restored speech to mix in the background.
Source: kerouacsamerica (Kerouac's America, SUNY Potsdam), CC BY 3.0, via Freesound.
Each excerpt is level-matched for fair listening and otherwise unprocessed. Sources and licenses are listed under each recording and in the media credits.
What Diffio restores
A voice isolator for recorded speech has one job: keep the person talking and move everything else out of the way. Diffio does that with speech restoration, so it works on the file you already have rather than at the mic.
Traffic, crowds and busy rooms
Buses, horns, HVAC, chatter in a lobby and the roar of an arena sit underneath the voice. Diffio removes background noise like this from the speech and keeps it on a separate background track. For steady hum and hiss, see how Diffio removes background noise from audio.
Wind on the microphone
Field recordings outdoors catch gusts that hit the capsule directly. Diffio isolates the voice from that rumble in supported cases. Pro handles hard wind best.
Music under speech
A music bed under a narrator, a band playing behind an interview at an event, a radio in the room: the music goes to the background track, and you get the speech without it. Nothing is deleted, so you can bring the music back at any level.
Echo and distortion
Isolating a voice often means fixing the room too. Diffio also reduces room echo and repairs supported clipping and distortion in the same pass. Here's more on removing echo from recorded speech.
How much background comes back is up to you
Every result has 2 outputs, the restored speech and the background. Sometimes you want the applause, the laughter or a bit of the street, because it tells the listener where you were. The Background amount slider sets how much of it comes back, from fully clean to the full original background.
How it works in 3 steps
Upload the recording
Create a project and add the file: MP3, WAV, M4A, AAC, FLAC or OGG audio, or MP4, MOV, AVI, MKV or WebM video, up to 90 minutes and 2 GB. Pick Diffio 4.5 Flash for speed or Diffio 4.5 Pro for the hardest recordings.
Diffio restores the speech
Flash runs about 35× faster than real time and Pro about 10× faster. Pro restores every window of audio 3 times and keeps the cleanest result. With video, the picture stays exactly as it was.
Compare, mix and download
Switch between original and restored speech on the same playhead, set the Background amount and read the transcript. On a paid plan, download the restored speech, the background on its own or your mix, as audio or as the full video.
Watch the 36-second demo
Real app screens and the real audio from the 2011 NHL Draft scrum above: upload, restore, compare and mix.
Who it's for
- Interviewers and podcasters whose guest sat in a café or a kitchen. Interview audio cleanup covers the two-person case in more detail, and there's a page for podcast episodes.
- Video producers and YouTubers with on-location footage. Diffio restores the soundtrack and leaves the picture alone. See removing background noise from video.
- Field reporters and newsrooms recording scrums, pressers and street interviews where you can't control the room. Keep the mics; fix the sound after.
- Educators and event teams with lectures, panels and Q&A recorded in halls full of people. Read about restoring lecture recordings.
- Developers adding speech isolation to their own product. The same models run behind the Diffio REST API, with Python and Node SDKs and per-second billing.
Diffio vs manual tools
Each tool here is good at its own job. Manual editors such as iZotope RX are the choice for surgical, hand-tuned repair. Real-time filters such as Krisp are built for live calls. Diffio is for isolating speech in a recording you already have, quickly, with the background kept.
| Compared | Diffio | Manual editor (for example iZotope RX) | Live call filter (for example Krisp) |
|---|---|---|---|
| Best at | Isolating speech in existing recordings | Hand-tuned, surgical repair | Noise during calls and meetings |
| When it works | After the recording | After the recording | Live, while you talk |
| What you do | Upload, then listen | Select regions and adjust each process | Turn it on before the call |
| What you get | Restored speech and background as 2 tracks | One processed file | A filtered live microphone |
| Keep some background | Background amount slider | By blending layers yourself | Not applicable |
| Video files | Upload the video; picture unchanged | Export audio, then re-sync | Not applicable |
For a free quick pass on short audio files, Adobe Podcast Enhance Speech works well; its free tier takes files up to 30 minutes, audio only (Adobe plans page, October 2026). Our Adobe Podcast comparison has the details.
Pricing and limits
Every plan, Free included, accepts audio and video files up to 90 minutes and 2 GB. Prices checked against the pricing page on October 19, 2026.
Free
Try your own recording first.
- 3 uploads, audio or video
- 2-minute preview of each result
- Diffio 4.5 Flash
- Full playback and downloads on paid plans
Basic
6 hours per month for most workflows.
- 6 hours of audio or video per month
- Diffio 4.5 Flash and Diffio 4.5 Pro
- Full playback and downloads
- Restored speech, background or mix
Pro
30 hours per month for regular production.
- 30 hours of audio or video per month
- Diffio 4.5 Flash and Diffio 4.5 Pro
- Full playback and downloads
- Restored speech, background or mix
Building it into your own product? The API is billed per second, about $5 per hour on Flash and about $10 per hour on Pro, and starts with $5 of free credit. See API pricing.
Voice isolator questions
Can I isolate vocals from a song?
Diffio is built for recorded speech: interviews, talks, calls, field recordings and narration. To pull a singer out of a finished song for karaoke or a remix, a music stem separator is the right tool. When the music is a bed under someone talking, Diffio is a good fit: the speech comes back restored and the music lands on the background track.
What happens to the background noise?
It isn't thrown away. Diffio separates it into its own background track. Leave it out for clean speech, or raise the Background amount slider to bring back applause, laughter or room tone. On a paid plan you can download the restored speech, the background or the current mix.
Can I isolate a voice from a video?
Yes, on every plan, including Free. Upload MP4, MOV, AVI, MKV or WebM. Diffio restores the soundtrack and leaves the picture as it was, so there is nothing to re-sync. Paid plans download the full video with the restored audio.
Can Diffio separate 2 people talking at once?
Diffio keeps every close voice together on the restored speech track. It doesn't split speakers into separate tracks, so overlapping talkers stay overlapped. A voice far from the mic, like a reporter shouting a question from the back of a scrum, can end up on the background track instead.
Does it work on a live call or a microphone?
Diffio works on recordings after they're made. You upload a file and get the result back in minutes. For noise during a live call, a real-time filter such as Krisp is built for that job.
Should I use Diffio 4.5 Flash or Diffio 4.5 Pro?
Flash is fast, about 35× faster than real time, and is on every plan. Pro is the best quality and runs about 10× faster than real time, because it restores every window of audio 3 times and keeps the cleanest result. Use Pro for wind, crowds and loud rooms. The Pro model is included in the Basic plan.
Is the voice isolator free?
The Free plan includes 3 uploads, audio or video, with Diffio 4.5 Flash and a 2-minute preview of each result. Full playback and downloads require a paid plan. Basic is $8 per month for 6 hours, and Pro is $29 per month for 30 hours.
What files and lengths can I upload?
Audio in MP3, WAV, M4A, AAC, FLAC or OGG and video in MP4, MOV, AVI, MKV or WebM. Each file can be up to 90 minutes long and up to 2 GB, on every plan.
Which languages does it support?
Both models are multilingual. They have been validated on English, French, German, Spanish and Portuguese.
What if the voice is far from the mic or badly distorted?
Results depend on the recording. A speaker across the room, heavy clipping and speech buried under louder speech are the hardest cases, and some words may stay rough. Upload the clip on the free plan and compare the restored speech against the original before you decide.
When the recording is all you have
Upload your toughest clip. You'll hear the original, the restored speech and the background side by side before you decide.