Rescript vs Cleanvoice: free filler removal, no upload

Rescript vs Cleanvoice

Cleanvoice is a service, not an editor: upload audio, its models strip filler words, stutters, mouth sounds, and dead air, and you download a cleaned file, billed per hour. Rescript is an editor that does the filler and silence removal locally, in one click each, for free, and then lets you keep editing by transcript. Cleanvoice detects more categories of nuisance — mouth sounds and stutters especially. Rescript costs nothing, never uploads your audio, and leaves you in control of every cut it makes.

Try it in your browser Download desktop app Source

The Rescript editor: a transcript on the left with a struck-through word, video preview on the right, and a waveform timeline underneath.

Every removal is a visible, reversible region rather than a processed file handed back to you. Nothing is decided off-screen.

Side by side

Rescript compared with Cleanvoice.

Feature comparison between Rescript and Cleanvoice
Feature	Rescript	Cleanvoice
Price	Free and open source for noncommercial use	Paid per hour of audio, or by subscription
Where your media goes	Nowhere. It never leaves your device	Audio is uploaded to be processed
Account required	None — open it and start editing	Required
Works offline	Yes, once the model has downloaded a first time	Cloud service
Source code	Public on GitHub — auditable	Proprietary
Filler word removal	One click across the whole file	Fillers, plus stutters and mouth sounds
Silence removal	One click, pauses of 0.3s and over	Dead air removal
Delete words to cut footage	Word-level, with a live preview of the cut	Not an editor — it returns a processed file
Noise removal / audio repair	No — nothing like Studio Sound	Mouth-sound removal, de-essing, levelling
Rewrite a line and have it spoken	Yes — cloned from that speaker's own audio, on-device	No
Timeline	Waveform, split, trim handles, draggable cut edges	No editing interface

Details about other products come from their own public pricing and documentation pages and change often — check theirs before you decide. Corrections are welcome as a GitHub issue.

A service versus an editor

Cleanvoice is designed to disappear into your workflow: send audio, get better audio. Its detection covers things Rescript has no concept of — mouth sounds, stutters, repeated words. If those are what make your raw audio unusable, Rescript won't fix them.

Rescript's filler removal is a click inside an editor. Every cut lands as a visible, editable region: you can see which words went, restore any of them, and drag a cut's edges. Nothing is billed by the hour.

The upload question

Per-hour cloud processing means every episode gets copied to a third party. For a public podcast that's usually a non-issue. For a research interview, a legal recording, or anything under NDA, it can be disqualifying regardless of the vendor's policies.

Rescript's processing is local by construction — Whisper in your browser, diarization locally, ffmpeg.wasm on your CPU. After the models download once, you can work with the network off.

Which one

They're good at different jobs.

Choose Cleanvoice if…

Mouth clicks, lip smacks, and stutters are the actual problem, not just "um".
You want it to run as a step in an automated pipeline via their API.
You'd rather not review the cuts and just want a cleaner file back.
You need loudness normalisation and de-essing in the same pass.

Choose Rescript if…

You don't want to pay per hour, and you record a lot of hours.
The audio can't be uploaded to a third-party service.
You want to see and approve each cut, and adjust the ones you disagree with.
You also need to cut content, not just clean it.

FAQ

Questions people actually ask

Is there a free alternative to Cleanvoice?

Rescript removes filler words and silences from audio and video on your own machine, free for noncommercial use and with no per-hour billing. It does not remove mouth sounds or stutters, which Cleanvoice does.

Does Rescript remove breaths and mouth sounds?

No. It detects filler words and silences of 0.3 seconds and over. Breaths, lip smacks, and stutters aren't detected — you can cut them by hand on the timeline, but they aren't automated.

Which filler words does Rescript detect?

Common disfluencies such as um, uh, er, and similar, in the transcript language. Detection runs against the local Whisper transcript, so anything it misses can be deleted by selecting the word.

Keep comparing