Seston is built around the work radio and podcast actually involve: long takes, a great deal of cutting, and getting a voice to sound like it was recorded somewhere better than it was.
Installs under your own account. No administrator prompt.
Enhance Speech
Most voice recordings have the same handful of things wrong with them: a noise floor, a room, a microphone with a tilt to it, and a level that wanders. Seston measures all of that in the file itself and fixes each one — in an order that matters.
The room comes off before the tone is measured, because reverberation has its own spectrum and would otherwise be corrected as though it were the microphone. The tone is matched before the level is evened out, because a compressor reacts to whatever is loudest and that ought to be a voice rather than a resonance. Loudness is set last, after everything that changes it.
Noise and reverberation are removed by a speech model that has been shown a great deal of clean speech, so it knows what belongs in a voice. It runs on your own machine.
Measured to ITU-R BS.1770 and set to −16 LUFS for podcast, −14 for streaming, −23 for broadcast. The limiter looks ahead, so peaks are limited rather than clipped.
Every measurement and every step is reported: the room's decay time, how far the tone was out, where the pauses were shut, the gain applied.
Editing
Smart Edit finds the dead air in a long take, takes it out, and lays what is left across two tracks head-to-tail, alternating — each join overlapping its neighbour by a crossfade.
That alternation is the point: butted cuts on a single track have nowhere to fade to. It is how somebody editing by hand would lay it out, and once it is right, Merge Clips renders the pieces back down to one — crossfades and all.
Trimming changes a clip's window onto the source, never the audio. The file you imported is never written to.
Drawn across each clip by default. Add handles and drag them to shape the level. The line belongs to the clip, so it travels with it.
Magnetic to the grid, to clip edges and to the playhead — with a tolerance capped in milliseconds, so zooming out does not turn snapping into guessing.
Double-click a clip to work on it alone: select a range and pull its level, fade from the edges, profile the room noise from a pause.
Recording
Set the source to System playback and Seston records whatever the machine is playing — the other person on a phone, WhatsApp or Zoom call — without touching the call application at all.
No link to send, no service to sign up for, and it works with anything that makes sound. Loopback only hears what the machine plays, which on a call is them and not you, so your microphone is captured at the same time onto its own track. The two sides land separately and can be edited apart.
Mixing and effects
High-pass, low-pass, EQ, gate, compressor, de-esser and limiter — on a clip, a track or the master, in a window you can leave open while you listen.
Eleven chains for voice come built in and anything you set up can be saved beside them. The signal path is fixed and worth knowing: clip source, clip gain and fades, clip effects, summed into the track, track effects, pan and fader, summed into the master, master effects, master gain, safety limiter.
Download
Windows 10 or 11, 64-bit. No account, no subscription, nothing to sign up for.
Release notes
Every version and its notes are on the releases page.
Seston