Voice processing

Jive Vocals

Raw microphone recordings into broadcast-ready audio in one command. No configuration, and no surprises.

Formerly Jivetalking. First discussed on Linux Matters 74.

Your audio engineer in a box.

# Process every presenter in parallel
$ jive-vocals mark.flac popey.flac

# Check a recording without processing
$ jive-vocals --analysis-only mark.flac

# Write before and after spectrograms
$ jive-vocals --diagnostics mark.flac

Measured, not guessed

Broadcast loudness.
No dials to turn.

Your files emerge at -16 LUFS / -1 dBTP, the loudness standard for spoken-word podcasts.

Read how the pipeline works

Processing that fits each recording: Jive Vocals analyses each recording first. Measured speech level sets the compressor threshold, and measured sibilance controls the de-esser.
The clean-up, sorted automatically: Room rumble, background hiss, clicks, and harsh sibilance are sorted without a setting to choose.
Every presenter at once: Multiple files process in parallel, each with its own progress row. Two star ratings then score your recording and the processed result.

The four-pass pipeline

Analyse. Process. Measure. Normalise.

Four passes carry a raw recording to a broadcast-ready master.

Analyse

Measure loudness, noise floor, and speech. Detect the room tone.
Process

Run the adapted filter chain. Each stage hands the next a cleaner signal.
Measure

Read the processed signal back, so that normalisation has accurate numbers.
Normalise

Set the final loudness to -16 LUFS / -1 dBTP.