An interview can be usable and still sound tired: a fridge hum under every answer, a music bed that fights the speaker, or room tone that makes the edit feel flat. Audio Cleanup in SmoothyApp handles those jobs on your computer, without uploading the recording or spending Studio credits.
Choose a recording and SmoothyApp separates it into Speech and Music. Noise is not a third track: it stays with the music. Turn on Enhanced when you want only the cleaned voice, with music and noise removed.
What the two tracks do
Speech is the separated voice.
Music contains everything outside the speech: the music bed, room tone, hiss and other background noise.
In Original, both tracks play together at their exact levels, so the split sounds like the recording. There are no Less / More controls and no processing beyond the separation.
In Enhanced, SmoothyApp cleans the speech with the local UVR DeNoise Lite model. Only the enhanced speech plays and the music and noise are removed. The Less / More control appears here for the speech preview; the saved WAV keeps the enhanced speech at its full level.
If the recording has overlapping speakers, heavy reverb or loud sound effects, expect artifacts. Separation is not a repair tool for a broken recording.
Set up the local engine once
The first separation needs a one-time setup:
- Open Audio Cleanup in SmoothyApp.
- Choose a file.
- On first use, keep SmoothyApp open while it creates the local environment and downloads the models. Later audio uses the engine again without another download.
Setup needs Python 3.12 or uv, an internet connection and several hundred megabytes of disk space. After it finishes, processing works offline. The engine runs on the CPU, so a long interview can take several minutes.
If setup fails, the app explains what is missing. Install Python 3.12 or uv, then use Repair engine & retry.
Process a recording
- Click Choose file and pick an audio or video file. Video files are supported; Smoothy decodes the first audio stream.
- Wait for the progress panel. SmoothyApp separates Speech and Music as soon as the file is ready. You can cancel while it runs.
- Listen to Original to check the exact split. The two tracks play together unchanged.
- Turn on Enhanced to clean the speech and remove music and noise. Switching between Original and Enhanced reuses the finished separation; it does not process the file again.
The source file is never modified. The source is decoded to stereo 44.1 kHz before processing, and each result is saved as a separate WAV.
Review before you save
Listen on headphones, then on speakers. A little room tone is often more natural than absolute silence, and aggressive noise removal can thin out consonants and breaths.
Check these moments specifically:
- the first and last words of each answer
- names and numbers
- laughter and overlapping speech
- the transitions between speakers
- any section where the music bed changes
If the enhanced voice sounds hollow or watery, compare it with Original. The exact split keeps the room and music, so you can save it and rebuild the mix manually. Enhanced is the right choice when the voice must stand alone.
Save and bring it into your edit
In Original, click the save button on Speech or Music to keep the exact split. The Music WAV includes the room tone and noise.
In Enhanced, click save on the speech card to keep the cleaned voice. Music and noise are not part of this result.
Audio Cleanup does not place the files on a timeline automatically. In Premiere Pro or DaVinci Resolve, import the WAVs and place them on your audio tracks. Keep the original recording in the project so you can compare or return to it. If you use Smoothy’s multicam or silence removal, a clean voice track can also make speaker detection more reliable.
Common questions
Does it upload my audio? No. Decoding, model processing and WAV output stay on your computer.
Do I need an account or credits? No. Audio Cleanup is a free local tool. Only the initial engine and model download needs internet.
Can I keep the removed noise? Yes, in Original. Noise stays with the music, so save the Music WAV to keep it.
Can I change the music level? Not in the exact split. Enhanced removes the music completely. Less / More applies only to the enhanced speech preview.
Why is only one audio stream used? The tool decodes the first audio stream. If your file has separate tracks for each speaker, export or prepare the one you need first.
Can it remove wind, traffic or a single cough? It is designed for steady noise. Intermittent sounds and complex mixes are better handled with manual editing.
Explore Audio Cleanup, silence removal and the sound design guide. For a full workflow, see recording separate tracks.