Uploads

client.process()

client.process() is the main entry point of the SDK. Pass it your audio or video source and any editing options — it handles upload (when needed), job submission, polling, and returns the result when processing is complete.

result = client.process(source, fillers=True, normalize=True)

Arguments: source — your file input (see options below) · editing options as keyword args (see Configuration)

Returns: a result object with result.audio.url and result.download_audio()

Overview — which input to use

Option 1: Public URL

What to pass: a public http:// or https:// URL

How it works: Cleanvoice fetches the file directly from your URL — no upload step on your end. The URL must be publicly accessible (no auth headers).

result = client.process(
    "https://example.com/episode.mp3",
    fillers=True,
    normalize=True,
)

Works with S3, GCS, Dropbox (?dl=1), Google Drive, or any CDN direct-download URL.

Option 2: Local file path

What to pass: a local file path as a string

How it works: The SDK detects a local path, uploads the file to Cleanvoice's storage automatically, then submits the edit job — all in one call. You get back the same result object as with a URL.

result = client.process(
    "/path/to/episode.mp3",
    fillers=True,
    normalize=True,
)

Option 3: NumPy array

What to pass: a (audio_array, sample_rate) tuple

How it works: The SDK encodes the array in memory and uploads it, skipping any temporary file. Useful in Jupyter notebooks, librosa pipelines, or anywhere audio is already decoded.

from cleanvoice import Cleanvoice

client = Cleanvoice.from_env()

audio_array, sample_rate = load_audio("episode.wav")  # your loading logic

result = client.process(
    (audio_array, sample_rate),
    fillers=True,
    normalize=True,
)

# Download result as a NumPy array
cleaned_audio, out_sample_rate = result.download_audio(as_numpy=True)

Bonus: Upload separately with upload_file()

What to pass: a local file path to upload_file() — then use the returned URL in client.process()

How it works: Uploads the file once and gives you a hosted URL you can reuse across multiple process() calls (different configs, retries, A/B). Useful when you want to separate ingest from job submission.

remote_url = client.upload_file(
    "/path/to/episode.mp3",
    filename="episode.mp3",  # optional
)

result = client.process(remote_url, fillers=True, normalize=True)

Supported formats

For local uploads, use the formats above. Remote URLs can still point to other media files that the Cleanvoice API can fetch directly.