These tools fix or take apart a track you already have. They run as AI generations from the Audio tab and cost a fraction of a cent at editor lengths. Reach for them early rather than living with a rough recording.
Open the Audio tab and choose Rework a track On the Generate tab, switch the output type to Sound and pick "Rework a track". Unlike the other sound options, this one takes an audio track as its input rather than a prompt.
Choose the track Pick any audio asset from your library in the track dropdown: an uploaded recording, a generated voiceover, or the audio you lifted off a video clip.
Pick what to do with it The model list is the menu of jobs. Denoise, isolate the voice from its background, or split out one stem (vocals, drums, bass, guitar or piano). Or extend the track, patch a bad section, or change the voice while keeping the delivery.
Set the specifics Stem splitting asks which part to keep. Extending asks how much to add and in what style. Patching asks for the start and end of the section to regenerate, on two sliders in seconds.
Drop it on a lane The reworked track lands in your library as a new asset. Drag it onto an audio lane and mix it like anything else; the original stays as it was.
Tips
For a lightly noisy clip, try the free export-time cleanup first: loudness normalisation and noise reduction are per-clip switches that cost nothing.
To use a song as a bed under narration, split out "everything else", which is the backing without the lead vocal.
A music track too short for the edit does not need looping: extending it continues the same style for up to a minute more.
Voice changing keeps the timing of the original performance, so a swapped voice still fits the picture and the captions.