New: Studio Sound v4

Podcast and video editing APIEdit recorded speech with one API call.

Automatic podcast and video editing from code: remove filler words, long silences, stutters, mouth sounds and breaths, in 20+ languages. Python and JavaScript SDKs, or REST.

Um, so today we're like, going to uh, clean up you know, the episode.

Tap for sound
Tap for sound
Tap for sound
Tap for sound

Or drop a file here, or .

Record up to 15 seconds, or drop audio or video up to 20 MB. Free, no account needed.

Read the docs
Veed
bHuman
Altered
Thames College
Teleprompter
Natural Reader
Veed
bHuman
Altered
Thames College
Teleprompter
Natural Reader

Remove filler words, silences and stutters from audio and video.

Every edit is a flag on the same request. Turn on the ones your recordings need; the rest of the file stays as it was.

  • Filler words

    Ums, uhs and other fillers, found in the speech itself, in 20+ languages.

    fillers

  • Long silences

    Dead air shortened, natural pauses kept.

    long_silences

  • Stutters

    Repeated words and false starts cut.

    stutters

  • Mouth sounds

    Clicks, lip smacks and swallows removed.

    mouth_sounds

  • Breaths

    Cut, softened or muted, per request.

    breath

  • Transcript

    A timestamped transcript of the edited file.

    transcription

Read the spec sheet.

What a developer usually scrolls the docs for, before deciding if an API is worth a key.

Every option in the docs →
Endpoint
POST api.cleanvoice.ai/v2/editsSend your key in the X-API-Key header, then poll GET /v2/edits/{id} until the file is done.
Editing flags
fillers · long_silences · stutters · mouth_sounds · breathBooleans on the same request. Add muted: true to silence the edits instead of cutting them, so the timing stays.
Languages
20+ languagesFiller words and stutters are detected in each of them.
Batch and multitrack
Many files, many speakersSend a batch of separate files, or a multitrack recording with one speaker per track.
Timeline export
The edits as a timelineExport where each cut was made and finish the edit in your own editor.
Processing speed
5 to 10 min per hour of audioA 2 to 3 minute clip is usually back in about 30 seconds.

Pick the edits. The code follows.

Turn on what your recordings need. The request changes with it, in Python, JavaScript or cURL.

Every option in the docs →
1from cleanvoice import Cleanvoice
2
3client = Cleanvoice.from_env()
4
5result = client.process(
6    "raw-take.wav",
7    fillers=True,
8    stutters=True,
9    mouth_sounds=True,
10    long_silences=True,
11)
12
13print(result.audio.url)
$ pip install cleanvoice-sdk

Pricing

From $0.01/ minute of audio

That's $0.60 an hour at enterprise volume, billed monthly or yearly. Self-serve plans from $0.75 an hour, or pay as you go at $1.50.

Compare plans
200 free hours

Startups get up to 200 hours of API credit. Request it from the API Keys page in your dashboard.

Request credits

Why developers trust us

All your data in your control. Each file processed securely.

  • EU-Hosted

    Servers and GPU located in the EU

  • ISO 27001 & GDPR Compliant

    DPA, SLA and on request even on premise deployment

  • No Training on Your Data.

    We record our own dataset in professional studios.

Questions about the audio editing API

Still have a question?

Send it over and a real person will answer.

support@cleanvoice.ai
Ask Now

Filler words such as um and uh, long silences, stutters and false starts, mouth sounds like clicks and lip smacks, and breaths. Each one is a flag, so you choose per request what goes.

Yes, it edits speech in 20+ languages, including speakers with an accent.

Set muted: true and the edits are silenced instead of cut out. The file keeps its exact length, which matters when the audio has to stay in sync with video or a music bed.

Yes. A batch is handled as separate files. A multitrack recording, one speaker per track, is edited as one conversation.

Yes. Send an MP4, MOV or MKV with long_silences on, and the edited video comes back with the silent parts cut from picture and sound together.

auto-editor is an open-source command-line tool that cuts the quiet parts of a file by loudness, on your own machine. It works well for dead air. The Cleanvoice API listens to the speech itself, so it also finds filler words, stutters, mouth sounds and breaths, and it is a hosted API your product calls, with nothing to install.

ffmpeg's silenceremove filter and pydub's split_on_silence cut wherever the level drops under a threshold you set. That handles dead air in a clean recording. Filler words and stutters are speech at normal volume, so a level threshold cannot find them; the API detects them in the words and cuts them in the same pass as the silences.

Yes. Export the timeline of edited parts and finish the cut in the editor you already use.

Yes. Add studio_sound for Studio Sound v4, our speech enhancement model, or remove_noise and normalize on their own. One file comes back with every edit applied.

About 5 to 10 minutes for an hour of audio, and around 30 seconds for a 2 to 3 minute clip.

Audio and video: .mp3, .wav, .m4a, .flac, .mp4, .mov and .mkv. A video comes back as video.

By the length of the audio, rounded up to the minute, pay as you go or with a subscription, at the same price as the app. You are not charged if a job fails. Startups can request up to 200 hours of free API credit from the API Keys page in the app.

Original and edited files are kept for 7 days, then deleted permanently. Cleanvoice is EU-hosted, GDPR compliant and ISO 27001 certified.

Get your API key.

Sign up, copy your key and send your first file.

Get your API keyBook a call