If the job is improving an existing recording, start with ElevenLabs Audio Isolation.
That is still the clearest cleanup-first path in Rivya.
This Page Is About Repairing Existing Audio
This guide follows Rivya's live audio catalog as it stood on April 21, 2026.
public paths cross-checked:
/audio,/ai-models, and current live audio model pagesrelated product guides reviewed: Audio Workflows in Rivya, References and Uploads in Rivya, and Current Live Features in Rivya
this page is only about improving an existing recording, not generating new speech, effects, or music
The first useful question here is very simple:
are you fixing something you already have?
or are you trying to create something new?
Those are different jobs and should leave through different doors.
Choose By Source Asset, Not By The Word "Audio"
Use ElevenLabs Audio Isolation when the source file already exists and the goal is to make it more usable.
That is the better path for:
cleaning a spoken recording
isolating vocals
preparing an audio file for editing, subtitles, dubbing, or reuse
If the real job is generating a new effect from text, switch to ElevenLabs Sound Effect V2.
If the real job is generating new speech, switch to the spoken-voice pages instead of staying here.
Why This Path Is Different
Cleanup work behaves differently from prompt-driven generation:
it starts from an upload
it depends on the quality of the source file
the result is judged by improvement, not invention
the cost pattern follows audio length more directly
That is why cleanup should be treated as its own workflow branch.
A Faster Cleanup Decision Order
If you want the shortest reliable order, use this:
decide whether the task is repair or generation
if it is repair, sign in before the upload step
upload the source file and stay in the cleanup path
only switch to prompt-driven models when the real job becomes new generation
That avoids the most common bad first move here: forcing a generator to do a repair job.
Where To Go Next
If the real task is generating a new sound effect, read Best AI Sound Effect Generator in 2026.
If the real task is new speech generation, read Best AI Voice Generator in 2026.
If you need the broader workflow guides, read Audio Workflows in Rivya and References and Uploads in Rivya.
Prepare A Cleanup Brief
Cleanup work starts from the source file, so the brief should describe the repair job instead of inventing a new sound.
Write down:
source type: interview, voice memo, screen recording, vocal take, or production audio
main problem: background noise, weak separation, rough capture, or reuse preparation
what must stay intact, especially timing and spoken content
target use: editing, subtitles, dubbing prep, video reuse, or archive cleanup
acceptable tradeoff between noise reduction and natural sound
whether the output should be downloaded immediately or kept for another workflow
That keeps the cleanup path separate from sound-effect or voice generation. If the source file is the asset, the task is repair first.
Review The Cleaned Audio
Judge the result by improvement, not by novelty.
Check:
whether speech or the target sound is clearer
whether timing stayed aligned with the original file
whether noise reduction introduced metallic artifacts
whether the output still feels natural enough for the target use
whether the source is too damaged and needs a different plan
whether reuse rights depend on the uploaded recording itself
If the cleaned file is useful, save it in History before using it downstream. If it is not useful, review the source and cleanup goal before switching to a generation model.
