Skip to content

Releases: EmptyOverlord/hush

Hush 1.1.0

Choose a tag to compare

@github-actions github-actions released this 06 Sep 05:53

The engine is now auto-editor 31.5.0 (was 29.3.1), and everything it
gained since is wired into the window.

New in the cutting

  • Cut by sound, by motion, or by either. Motion keeps screen recordings
    where the picture moves and nobody is talking — those used to be thrown away
    as silence.
  • Speed up the talking itself, up to ×2, separately from what happens to
    the pauses.
  • Smoothing merges micro-cuts, so speech stops stuttering at the seams.
  • Drop black frames — fades, title cards, a covered lens.
  • Smooth transitions: a dissolve across every cut instead of a hard join.
    Works for finished video and for editor timelines alike.

Subtitles, offline

Hush now transcribes. Recognition runs on your machine — nothing is
uploaded, and once the model is downloaded it works with the network off.

  • SRT to upload straight to YouTube, or plain text
  • One word per line — the bouncing captions Shorts are made of
  • Translate into English in the same pass: record in your language, get
    English subtitles out
  • Eight models, from tiny (74 MB) to large-v3 (2.9 GB). The quantised ones
    give you a big model at a fraction of the weight — turbo-q5 (547 MB) is
    the sweet spot. Downloaded models are marked, so you can see what costs a
    download before you pick one.

Fixes

  • --temp-dir is gone from the engine as of 31.1.0. Hush was still passing
    it, so pressing Update engine would have broken every run afterwards. It
    now asks the engine what it understands.
  • Release assets were renamed upstream (windows-amd64 → windows-x86_64), so
    every known name is tried when downloading. ARM builds for Windows and
    Linux are picked up too.
  • Settings no longer clip: tab bodies scroll when they outgrow the panel, and
    captions wrap to the real panel width instead of a hardcoded one.
  • Switching language keeps you on the tab you were on.
  • A transition longer than the clips between cuts now says so in plain words,
    instead of surfacing a raw engine error.

Interface

Settings moved onto four tabs — Cutting · Fine-tune · Output · Subtitles —
because two could not hold this. Every slider and every checkbox now says what
it does underneath. Opening the Subtitles tab moves the primary button onto
Make subtitles, since transcribing is a separate job from cutting.

Hush 1.0.0

Choose a tag to compare

@github-actions github-actions released this 06 Sep 03:14

Hush cuts the pauses out of your video. Point it at footage of yourself
talking, and it removes every gap where you were breathing, thinking, or
reaching for the mouse.

The auto-editor engine is bundled
inside every build. Nothing is downloaded at first run, nothing gets installed,
and it works offline.

Pick your download

File For
Hush-macOS-arm64.zip Mac with Apple Silicon (M1–M4)
Hush-macOS-x86_64.zip Intel Mac
Hush-Windows-x64.zip Windows 10/11, 64-bit
Hush-Linux-x86_64.zip Linux, 64-bit

macOS: the app is not notarised, so Gatekeeper will complain the first time.
Right-click Hush.app → Open → Open. Once, and never again.

In this release

  • Drag and drop, a file queue, and three cutting presets
  • Padding and threshold sliders — tune it to your room and your voice
  • Cut the silence, or speed it up ×2 / ×4 instead
  • Count without rendering: see exactly how much will go, in about a second
  • Export a finished video, or a ready-cut timeline for DaVinci Resolve,
    Premiere Pro, Final Cut Pro, Shotcut or Kdenlive — nothing re-encoded
  • Loudness normalisation on video output
  • Russian and English, switched with one button
  • Engine updates from inside the app, when you're online

Your source files are safe

Output goes to a Hush folder next to the original. The app refuses to write to
the input path, so the file you fed it is never modified.

What it can't do

The engine measures loudness, not speech. Background music defeats it,
overlapping voices confuse it, and quiet consonants can get clipped — that's
what the padding slider is for. It is built for one person, one microphone, one
quiet room, and there it saves hours.