Repository navigation
Releases
VoiceCut 0.4.3
VoiceCut listens to a narration recording, keeps the best take of every line,
fixes the pauses, and lets you fine-tune every cut on a timeline before
exporting.
What's new in 0.4.3
- Pauses are back when breaths are removed. In a tightly spoken
recording, with a breath in every pause, removing the breaths could leave
almost no pauses at all (0.4.2 did this to Russian recordings). Every pause
now keeps its length: what the recording's clean quiet cannot fill gets
room tone from short quiet stretches of the same recording, or silence when
it has none. In a video, such a pause keeps its recorded sound instead.
New in 0.4.2
- Breaths removed in Russian too. Voice Cut now replaces breaths, clicks
and mouth noise in pauses with room tone in every language; Russian
recordings used to keep them. Settings › Voice Cut has a Remove breaths
checkbox, ticked by default: untick it to keep every pause as recorded. A
change applies the next time you run Voice Cut.
New in 0.4.1
- Sound after a long Voice Cut. On a Mac, the edit of a long recording
could play silent after Voice Cut finished until VoiceCut was restarted.
The player now rests its audio output while nothing plays and wakes it to
play, which brings the sound back.
New in 0.4.0
- Fast planners for smaller Macs. Kev 9B and Kev 4B, decision models that
answer typed questions about every line instead of writing text, plan an
hour of narration in about 10–20 minutes on Apple Silicon from 12 GB of
memory, a little less precisely than Gemma 4. New projects pick the best
planner your computer runs well: Gemma 4 31B, else Kev 9B, else Kev 4B,
else Gemma 4 through OpenRouter. TypeSafe Jev, a hosted decision model, is
in the OpenRouter list. - See how long it will take. Voice Cut shows how far each stage has got
and the time left, learned from how fast your own computer ran before. - Try another model without waiting again. Running Voice Cut again with
another planner reuses the transcription: only the take selection and the
render run again, and going back to a model you used is instant. - Your own pause lengths. Settings › Pauses sets how long Voice Cut's
pauses are — inside a sentence, after a comma, between sentences and
paragraphs — for every project at once. Pauses you resized keep their length. - The top bar fits narrow windows: it hides its text labels instead of
letting the controls overlap. The language is now chosen asenorru. - The “Back to Voice Cut's edit” button is gone: running Voice Cut again
with the same settings brings its edit back at once.
Install
| System | File | |
|---|---|---|
| macOS 14+, Apple Silicon or Intel | VoiceCut-<version>-universal.dmg |
Open it and drag VoiceCut to Applications. |
| Linux x86_64 (glibc 2.35+: Ubuntu 22.04, Debian 12, Fedora 36 or newer) | VoiceCut-<version>-x86_64.AppImage |
chmod +x it and run it. It adds itself to your applications. |
| Debian, Ubuntu | VoiceCut-<version>-amd64.deb |
sudo apt install ./VoiceCut-<version>-amd64.deb |
| Windows 10/11 x64 | VoiceCut-<version>-x64-setup.exe |
Run it; it installs for you alone, no administrator needed. |
The AppImage needs FUSE 2: if it does not start on Ubuntu, install it with
sudo apt install libfuse2 (libfuse2t64 on Ubuntu 24.04 and later).
On first launch VoiceCut downloads and sets up its engine (about 1.7 GB on a
Mac and 2.9 GB on Windows and Linux; keep up to 9 GB free while it installs).
The planner runs on your computer, or with your own API key for a cloud
provider (OpenRouter, Gemini, OpenAI, DeepSeek), which VoiceCut keeps in the
system's keychain or credential store.
Whisper runs on the GPU on Apple Silicon (MLX) and on NVIDIA GPUs with a
recent driver (CUDA 12); anywhere else it runs on the CPU, about as fast as
the recording plays.
Updates
VoiceCut updates itself: when a new version is out, a download button appears
in its toolbar, and unless you turn it off in Settings the update installs
when you quit VoiceCut. latest.json and the .app.tar.gz are what it
downloads; you don't need them. On Debian and Ubuntu the update asks for
your password.
Unsigned builds
Unless this release says it is signed, the installers are not signed with a
paid certificate yet:
- macOS says VoiceCut "cannot be opened because Apple cannot check it for
malicious software". Open System Settings → Privacy & Security and choose
Open Anyway (once). - Windows SmartScreen says it "protected your PC". Choose More info →
Run anyway.
VoiceCut 0.4.2
VoiceCut listens to a narration recording, keeps the best take of every line,
fixes the pauses, and lets you fine-tune every cut on a timeline before
exporting.
What's new in 0.4.2
- Breaths removed in Russian too. Voice Cut now replaces breaths, clicks
and mouth noise in pauses with room tone in every language; Russian
recordings used to keep them. Settings › Voice Cut has a Remove breaths
checkbox, ticked by default: untick it to keep every pause as recorded. A
change applies the next time you run Voice Cut.
New in 0.4.1
- Sound after a long Voice Cut. On a Mac, the edit of a long recording
could play silent after Voice Cut finished until VoiceCut was restarted.
The player now rests its audio output while nothing plays and wakes it to
play, which brings the sound back.
New in 0.4.0
- Fast planners for smaller Macs. Kev 9B and Kev 4B, decision models that
answer typed questions about every line instead of writing text, plan an
hour of narration in about 10–20 minutes on Apple Silicon from 12 GB of
memory, a little less precisely than Gemma 4. New projects pick the best
planner your computer runs well: Gemma 4 31B, else Kev 9B, else Kev 4B,
else Gemma 4 through OpenRouter. TypeSafe Jev, a hosted decision model, is
in the OpenRouter list. - See how long it will take. Voice Cut shows how far each stage has got
and the time left, learned from how fast your own computer ran before. - Try another model without waiting again. Running Voice Cut again with
another planner reuses the transcription: only the take selection and the
render run again, and going back to a model you used is instant. - Your own pause lengths. Settings › Pauses sets how long Voice Cut's
pauses are — inside a sentence, after a comma, between sentences and
paragraphs — for every project at once. Pauses you resized keep their length. - The top bar fits narrow windows: it hides its text labels instead of
letting the controls overlap. The language is now chosen asenorru. - The “Back to Voice Cut's edit” button is gone: running Voice Cut again
with the same settings brings its edit back at once.
Install
| System | File | |
|---|---|---|
| macOS 14+, Apple Silicon or Intel | VoiceCut-<version>-universal.dmg |
Open it and drag VoiceCut to Applications. |
| Linux x86_64 (glibc 2.35+: Ubuntu 22.04, Debian 12, Fedora 36 or newer) | VoiceCut-<version>-x86_64.AppImage |
chmod +x it and run it. It adds itself to your applications. |
| Debian, Ubuntu | VoiceCut-<version>-amd64.deb |
sudo apt install ./VoiceCut-<version>-amd64.deb |
| Windows 10/11 x64 | VoiceCut-<version>-x64-setup.exe |
Run it; it installs for you alone, no administrator needed. |
The AppImage needs FUSE 2: if it does not start on Ubuntu, install it with
sudo apt install libfuse2 (libfuse2t64 on Ubuntu 24.04 and later).
On first launch VoiceCut downloads and sets up its engine (about 1.7 GB on a
Mac and 2.9 GB on Windows and Linux; keep up to 9 GB free while it installs).
The planner runs on your computer, or with your own API key for a cloud
provider (OpenRouter, Gemini, OpenAI, DeepSeek), which VoiceCut keeps in the
system's keychain or credential store.
Whisper runs on the GPU on Apple Silicon (MLX) and on NVIDIA GPUs with a
recent driver (CUDA 12); anywhere else it runs on the CPU, about as fast as
the recording plays.
Updates
VoiceCut updates itself: when a new version is out, a download button appears
in its toolbar, and unless you turn it off in Settings the update installs
when you quit VoiceCut. latest.json and the .app.tar.gz are what it
downloads; you don't need them. On Debian and Ubuntu the update asks for
your password.
Unsigned builds
Unless this release says it is signed, the installers are not signed with a
paid certificate yet:
- macOS says VoiceCut "cannot be opened because Apple cannot check it for
malicious software". Open System Settings → Privacy & Security and choose
Open Anyway (once). - Windows SmartScreen says it "protected your PC". Choose More info →
Run anyway.
VoiceCut 0.4.1
VoiceCut listens to a narration recording, keeps the best take of every line,
fixes the pauses, and lets you fine-tune every cut on a timeline before
exporting.
What's new in 0.4.1
- Sound after a long Voice Cut. On a Mac, the edit of a long recording
could play silent after Voice Cut finished until VoiceCut was restarted.
The player now rests its audio output while nothing plays and wakes it to
play, which brings the sound back.
New in 0.4.0
- Fast planners for smaller Macs. Kev 9B and Kev 4B, decision models that
answer typed questions about every line instead of writing text, plan an
hour of narration in about 10–20 minutes on Apple Silicon from 12 GB of
memory, a little less precisely than Gemma 4. New projects pick the best
planner your computer runs well: Gemma 4 31B, else Kev 9B, else Kev 4B,
else Gemma 4 through OpenRouter. TypeSafe Jev, a hosted decision model, is
in the OpenRouter list. - See how long it will take. Voice Cut shows how far each stage has got
and the time left, learned from how fast your own computer ran before. - Try another model without waiting again. Running Voice Cut again with
another planner reuses the transcription: only the take selection and the
render run again, and going back to a model you used is instant. - Your own pause lengths. Settings › Pauses sets how long Voice Cut's
pauses are — inside a sentence, after a comma, between sentences and
paragraphs — for every project at once. Pauses you resized keep their length. - The top bar fits narrow windows: it hides its text labels instead of
letting the controls overlap. The language is now chosen asenorru. - The “Back to Voice Cut's edit” button is gone: running Voice Cut again
with the same settings brings its edit back at once.
Install
| System | File | |
|---|---|---|
| macOS 14+, Apple Silicon or Intel | VoiceCut-<version>-universal.dmg |
Open it and drag VoiceCut to Applications. |
| Linux x86_64 (glibc 2.35+: Ubuntu 22.04, Debian 12, Fedora 36 or newer) | VoiceCut-<version>-x86_64.AppImage |
chmod +x it and run it. It adds itself to your applications. |
| Debian, Ubuntu | VoiceCut-<version>-amd64.deb |
sudo apt install ./VoiceCut-<version>-amd64.deb |
| Windows 10/11 x64 | VoiceCut-<version>-x64-setup.exe |
Run it; it installs for you alone, no administrator needed. |
The AppImage needs FUSE 2: if it does not start on Ubuntu, install it with
sudo apt install libfuse2 (libfuse2t64 on Ubuntu 24.04 and later).
On first launch VoiceCut downloads and sets up its engine (about 1.7 GB on a
Mac and 2.9 GB on Windows and Linux; keep up to 9 GB free while it installs).
The planner runs on your computer, or with your own API key for a cloud
provider (OpenRouter, Gemini, OpenAI, DeepSeek), which VoiceCut keeps in the
system's keychain or credential store.
Whisper runs on the GPU on Apple Silicon (MLX) and on NVIDIA GPUs with a
recent driver (CUDA 12); anywhere else it runs on the CPU, about as fast as
the recording plays.
Updates
VoiceCut updates itself: when a new version is out, a download button appears
in its toolbar, and unless you turn it off in Settings the update installs
when you quit VoiceCut. latest.json and the .app.tar.gz are what it
downloads; you don't need them. On Debian and Ubuntu the update asks for
your password.
Unsigned builds
Unless this release says it is signed, the installers are not signed with a
paid certificate yet:
- macOS says VoiceCut "cannot be opened because Apple cannot check it for
malicious software". Open System Settings → Privacy & Security and choose
Open Anyway (once). - Windows SmartScreen says it "protected your PC". Choose More info →
Run anyway.
VoiceCut 0.4.0
VoiceCut listens to a narration recording, keeps the best take of every line,
fixes the pauses, and lets you fine-tune every cut on a timeline before
exporting.
What's new in 0.4.0
- Fast planners for smaller Macs. Kev 9B and Kev 4B, decision models that
answer typed questions about every line instead of writing text, plan an
hour of narration in about 10–20 minutes on Apple Silicon from 12 GB of
memory, a little less precisely than Gemma 4. New projects pick the best
planner your computer runs well: Gemma 4 31B, else Kev 9B, else Kev 4B,
else Gemma 4 through OpenRouter. TypeSafe Jev, a hosted decision model, is
in the OpenRouter list. - See how long it will take. Voice Cut shows how far each stage has got
and the time left, learned from how fast your own computer ran before. - Try another model without waiting again. Running Voice Cut again with
another planner reuses the transcription: only the take selection and the
render run again, and going back to a model you used is instant. - Your own pause lengths. Settings › Pauses sets how long Voice Cut's
pauses are — inside a sentence, after a comma, between sentences and
paragraphs — for every project at once. Pauses you resized keep their length. - The top bar fits narrow windows: it hides its text labels instead of
letting the controls overlap. The language is now chosen asenorru. - The “Back to Voice Cut's edit” button is gone: running Voice Cut again
with the same settings brings its edit back at once.
Install
| System | File | |
|---|---|---|
| macOS 14+, Apple Silicon or Intel | VoiceCut-<version>-universal.dmg |
Open it and drag VoiceCut to Applications. |
| Linux x86_64 (glibc 2.35+: Ubuntu 22.04, Debian 12, Fedora 36 or newer) | VoiceCut-<version>-x86_64.AppImage |
chmod +x it and run it. It adds itself to your applications. |
| Debian, Ubuntu | VoiceCut-<version>-amd64.deb |
sudo apt install ./VoiceCut-<version>-amd64.deb |
| Windows 10/11 x64 | VoiceCut-<version>-x64-setup.exe |
Run it; it installs for you alone, no administrator needed. |
The AppImage needs FUSE 2: if it does not start on Ubuntu, install it with
sudo apt install libfuse2 (libfuse2t64 on Ubuntu 24.04 and later).
On first launch VoiceCut downloads and sets up its engine (about 1.7 GB on a
Mac and 2.9 GB on Windows and Linux; keep up to 9 GB free while it installs).
The planner runs on your computer, or with your own API key for a cloud
provider (OpenRouter, Gemini, OpenAI, DeepSeek), which VoiceCut keeps in the
system's keychain or credential store.
Whisper runs on the GPU on Apple Silicon (MLX) and on NVIDIA GPUs with a
recent driver (CUDA 12); anywhere else it runs on the CPU, about as fast as
the recording plays.
Updates
VoiceCut updates itself: when a new version is out, a download button appears
in its toolbar, and unless you turn it off in Settings the update installs
when you quit VoiceCut. latest.json and the .app.tar.gz are what it
downloads; you don't need them. On Debian and Ubuntu the update asks for
your password.
Unsigned builds
Unless this release says it is signed, the installers are not signed with a
paid certificate yet:
- macOS says VoiceCut "cannot be opened because Apple cannot check it for
malicious software". Open System Settings → Privacy & Security and choose
Open Anyway (once). - Windows SmartScreen says it "protected your PC". Choose More info →
Run anyway.
VoiceCut 0.3.1
VoiceCut listens to a narration recording, keeps the best take of every line,
fixes the pauses, and lets you fine-tune every cut on a timeline before
exporting.
What's new in 0.3.1
- Move and copy clips. Drag a clip to another place on the timeline;
hold ⌥ (Alt) to drop a copy instead. ⌘C, ⌘X and ⌘V copy, cut and paste
clips (a paste inside a clip splits it at the playhead), and ⌘D duplicates
the selection. Previews and exports, video included, follow the new order. - Select many clips at once. ⇧-click selects every clip and pause from
the last one clicked; ⌘-click (Ctrl-click on Windows and Linux) adds or
removes one; ⌘A selects everything. - Clicking a clip still puts the playhead there; dragging the ruler scrubs.
Install
| System | File | |
|---|---|---|
| macOS 14+, Apple Silicon or Intel | VoiceCut-<version>-universal.dmg |
Open it and drag VoiceCut to Applications. |
| Linux x86_64 (glibc 2.35+: Ubuntu 22.04, Debian 12, Fedora 36 or newer) | VoiceCut-<version>-x86_64.AppImage |
chmod +x it and run it. It adds itself to your applications. |
| Debian, Ubuntu | VoiceCut-<version>-amd64.deb |
sudo apt install ./VoiceCut-<version>-amd64.deb |
| Windows 10/11 x64 | VoiceCut-<version>-x64-setup.exe |
Run it; it installs for you alone, no administrator needed. |
The AppImage needs FUSE 2: if it does not start on Ubuntu, install it with
sudo apt install libfuse2 (libfuse2t64 on Ubuntu 24.04 and later).
On first launch VoiceCut downloads and sets up its engine (about 1.7 GB on a
Mac and 2.9 GB on Windows and Linux; keep up to 9 GB free while it installs).
The planner runs on your computer, or with your own API key for a cloud
provider (OpenRouter, Gemini, OpenAI, DeepSeek), which VoiceCut keeps in the
system's keychain or credential store.
Whisper runs on the GPU on Apple Silicon (MLX) and on NVIDIA GPUs with a
recent driver (CUDA 12); anywhere else it runs on the CPU, about as fast as
the recording plays.
Updates
VoiceCut updates itself: when a new version is out, a download button appears
in its toolbar, and unless you turn it off in Settings the update installs
when you quit VoiceCut. latest.json and the .app.tar.gz are what it
downloads; you don't need them. On Debian and Ubuntu the update asks for
your password.
Unsigned builds
Unless this release says it is signed, the installers are not signed with a
paid certificate yet:
- macOS says VoiceCut "cannot be opened because Apple cannot check it for
malicious software". Open System Settings → Privacy & Security and choose
Open Anyway (once). - Windows SmartScreen says it "protected your PC". Choose More info →
Run anyway.
VoiceCut 0.3.0
VoiceCut listens to a narration recording, keeps the best take of every line,
fixes the pauses, and lets you fine-tune every cut on a timeline before
exporting.
What's new in 0.3.0
- The planner can run on your computer. Local models run through Ollama
on macOS, Windows and Linux, and VoiceCut installs Ollama unless you have
it. New projects plan with Gemma 4 31B on your computer where it runs well
(Apple Silicon with 32 GB of memory or more, or an NVIDIA GPU with 24 GB),
and with the same model through OpenRouter elsewhere. The first Voice Cut
offers to download the model (about 20 GB, once) and starts when it is
there. - OpenRouter joins Gemini, OpenAI and DeepSeek. Gemma 4 31B there edits
nearly the way Gemini 3.1 Pro does, for about 5 cents of OpenRouter's
charges per hour of recording. Gemini now defaults to Gemini 3.8 Flash. - Voice Cut edits only what is left on the timeline. Trim a recording
first and only the parts you kept are transcribed and planned; delete part
of Voice Cut's result and the next run leaves it out too. - Intel Macs: the macOS download is now universal.
- Windows: Voice Cut no longer hangs at "Checking runtimes", and setup
works when Windows long paths are off. - The run's log opens at a readable size and says whether Whisper runs on
the GPU.
Install
| System | File | |
|---|---|---|
| macOS 14+, Apple Silicon or Intel | VoiceCut-<version>-universal.dmg |
Open it and drag VoiceCut to Applications. |
| Linux x86_64 (glibc 2.35+: Ubuntu 22.04, Debian 12, Fedora 36 or newer) | VoiceCut-<version>-x86_64.AppImage |
chmod +x it and run it. It adds itself to your applications. |
| Debian, Ubuntu | VoiceCut-<version>-amd64.deb |
sudo apt install ./VoiceCut-<version>-amd64.deb |
| Windows 10/11 x64 | VoiceCut-<version>-x64-setup.exe |
Run it; it installs for you alone, no administrator needed. |
The AppImage needs FUSE 2: if it does not start on Ubuntu, install it with
sudo apt install libfuse2 (libfuse2t64 on Ubuntu 24.04 and later).
On first launch VoiceCut downloads and sets up its engine (about 1.7 GB on a
Mac and 2.9 GB on Windows and Linux; keep up to 9 GB free while it installs).
The planner runs on your computer, or with your own API key for a cloud
provider (OpenRouter, Gemini, OpenAI, DeepSeek), which VoiceCut keeps in the
system's keychain or credential store.
Whisper runs on the GPU on Apple Silicon (MLX) and on NVIDIA GPUs with a
recent driver (CUDA 12); anywhere else it runs on the CPU, about as fast as
the recording plays.
Updates
VoiceCut updates itself: when a new version is out, a download button appears
in its toolbar, and unless you turn it off in Settings the update installs
when you quit VoiceCut. latest.json and the .app.tar.gz are what it
downloads; you don't need them. On Debian and Ubuntu the update asks for
your password.
Unsigned builds
Unless this release says it is signed, the installers are not signed with a
paid certificate yet:
- macOS says VoiceCut "cannot be opened because Apple cannot check it for
malicious software". Open System Settings → Privacy & Security and choose
Open Anyway (once). - Windows SmartScreen says it "protected your PC". Choose More info →
Run anyway.
VoiceCut 0.2.0
VoiceCut listens to a narration recording, keeps the best take of every line,
fixes the pauses, and lets you fine-tune every cut on a timeline before
exporting.
Install
| System | File | |
|---|---|---|
| macOS 14+ on Apple Silicon | VoiceCut-<version>-arm64.dmg |
Open it and drag VoiceCut to Applications. |
| Linux x86_64 (glibc 2.35+: Ubuntu 22.04, Debian 12, Fedora 36 or newer) | VoiceCut-<version>-x86_64.AppImage |
chmod +x it and run it. It adds itself to your applications. |
| Debian, Ubuntu | VoiceCut-<version>-amd64.deb |
sudo apt install ./VoiceCut-<version>-amd64.deb |
| Windows 10/11 x64 | VoiceCut-<version>-x64-setup.exe |
Run it; it installs for you alone, no administrator needed. |
The AppImage needs FUSE 2: if it does not start on Ubuntu, install it with
sudo apt install libfuse2 (libfuse2t64 on Ubuntu 24.04 and later).
On first launch VoiceCut downloads and sets up its engine (about 1.5 GB, up
to 7 GB of disk while installing). Voice Cut itself needs a Gemini API key,
which VoiceCut keeps in the system's keychain or credential store.
Whisper runs on the GPU on Apple Silicon (MLX) and on NVIDIA GPUs with a
recent driver (CUDA 12); anywhere else it runs on the CPU, about as fast as
the recording plays.
Updates
VoiceCut updates itself: when a new version is out, a download button appears
in its toolbar, and unless you turn it off in Settings the update installs
when you quit VoiceCut. latest.json and the .app.tar.gz are what it
downloads; you don't need them. On Debian and Ubuntu the update asks for
your password.
Unsigned builds
Unless this release says it is signed, the installers are not signed with a
paid certificate yet:
- macOS says VoiceCut "cannot be opened because Apple cannot check it for
malicious software". Open System Settings → Privacy & Security and choose
Open Anyway (once). - Windows SmartScreen says it "protected your PC". Choose More info →
Run anyway.