2.6.0

What's new

• Vocal Slice is now free and open source. There's no trial and no licence key any more — every feature is unlocked for everyone, and the full source is on GitHub under the GPL-3.0. If you bought a licence, thank you: you don't need to do anything, and nothing you had stops working.
• The License page has gone from Settings, along with the trial countdown in the tab bar. Vocal Slice also tidies away the licence and trial records it used to keep on your computer.
• Settings → About now shows the open-source licence notice and a link to the source code.
• Version numbers now start at 2. That's the move to the open-source repository, not a change in how updates work.
• Fixed the release notes in Settings → About showing raw formatting marks — **like this** — instead of bold text and code.

Which file?

• Windows — VocalSlice-Setup.exe
• macOS — VocalSlice.dmg, then drag Vocal Slice to Applications. VocalSlice-macOS-update.zip is the package the app updates itself from — not a manual download.

SHA-256 checksums

File	Size	SHA-256
VocalSlice-macOS-update.zip	198.5 MB	aa3e0661f7926911066d5a71feebc7a5f7722871990e9ce757e54313c1c578d6
VocalSlice-Setup.exe	94.4 MB	31f7b4fd520118f28d758b214cf55dd3e491b4a72f091f9b86a72081462eaad6
VocalSlice.dmg	205.5 MB	9898730853eff2ff78c3537e3f72ed63b4a4f60b60e85c319b7d3349b9258221

Verify: certutil -hashfile <file> SHA256 (Windows) · shasum -a 256 <file> (macOS)

---

Windows isn't code-signed yet, so SmartScreen warns on first run — choose More info → Run anyway; the checksum above is how you verify the download. macOS is signed and notarized.

1.475.0

What's new

• The first-run tour now points out that selecting a phrase finds every other place it appears, so you can step between takes of the same line instead of hunting for them.

Which file?

• Windows — VocalSlice-Setup.exe
• macOS — VocalSlice.dmg, then drag Vocal Slice to Applications. VocalSlice-macOS-update.zip is the package the app updates itself from — not a manual download.

SHA-256 checksums

File	Size	SHA-256
VocalSlice-macOS-update.zip	198.5 MB	047db5385c9281671d8b4b42337102244879189ae55d99881e15d05e206441d3
VocalSlice-Setup.exe	94.4 MB	62049c84929cfe0b1cd2a2178065f46b3976c848a380cfa0617fce571ebd9664
VocalSlice.dmg	205.5 MB	830e419e408e3a1f968a967f5102d9843b58c683a9481a04f124201ff83b0f4e

Verify: certutil -hashfile <file> SHA256 (Windows) · shasum -a 256 <file> (macOS)

---

Windows isn't code-signed yet, so SmartScreen warns on first run — choose More info → Run anyway; the checksum above is how you verify the download. macOS is signed and notarized.

1.474.0

What's new

• Vocal Slice is now a one-time purchase. Buy it once and it's yours — no subscription, no renewal, and every future version included. It's the same $29 it always was; you just don't pay it again.

• Once you've activated your key, Vocal Slice never needs to check in again. It used to re-confirm your subscription every couple of weeks, which meant a fortnight away from the internet could lock you out of your own software. That can't happen now: activate once and it keeps working, offline, indefinitely.

• Settings → License now tells you what it costs and what you get before you click through to the checkout, instead of sending you to a payment page to find out.

• The "Active until" line has gone from Settings → License. It never had a date to show, and a purchase that doesn't run out has no end date to report.

• Vocal Slice now tells you what it's doing in plain language while it gets started — "Downloading transcription model", "Reading your audio", "Transcribing" — instead of naming internal model files like decoder_model_merged_q4.onnx.

• The progress bar for a first-time model download now fills once, from start to finish. It used to restart from zero for each of the six files a model is made of, so it swept across six times and told you nothing about how much was actually left.

• If part of a model is already on your machine, the bar starts partway along rather than at zero, so it reflects what's genuinely left to download.

• Added Whisper Medium, in English-only and multilingual flavours — the most accurate model Vocal Slice offers, and by some way the slowest. It runs with or without a graphics card, though on a machine without one expect it to take several times longer than the audio itself; the smaller models remain much better company there.

• The model list is now grouped into English-only and multilingual, so it stays readable as it grows.

• Vocal Slice now asks before starting a model download over 1GB, and tells you how big it is, so a change of setting can't turn into a surprise 1.6GB download later on.

• Transcribing on your processor is now around two to three times faster. Vocal Slice was only ever using a single core for it, however many your machine has; it now uses half of them, leaving the rest free for whatever else you're doing. The transcript comes out identical — this is the same work spread wider, not a shortcut. Whisper Medium on a processor drops from roughly seven times the length of your audio to under three.

• Vocal Slice now warns you that it will stop responding while it transcribes on your processor, instead of simply freezing. Transcribing without a graphics card has always held the window still until the run finishes — on a long file with a large model that can be a good while, and with nothing on screen to say so it looked like a crash. Settings → Transcription says the same thing where you choose the processor.

• Transcription models now come from Vocal Slice itself rather than a third-party site. Previously the app downloaded them from Hugging Face, which meant that if those files ever moved or became unreachable, a fresh install had no way to transcribe anything. Nothing about your audio changes — it was always processed on your machine — but the app no longer depends on anyone else to work.

• Models are now saved as ordinary files in a folder you can open, back up, or copy to another machine. They used to live in hidden browser storage that nothing but the app could reach, and that "Clear Session Data" could wipe by accident.

• Every model file is checked against a known fingerprint as it downloads, so a partial or corrupted download is caught and fetched again instead of failing later with an unhelpful error.

• If our download servers can't be reached — some workplace networks block them — the app now quietly falls back to fetching the same model from Hugging Face, rather than leaving you unable to transcribe. The fingerprint check applies either way, so you always end up with the exact model we tested.

• The model picker now shows what each model will actually download on your machine. The old figures were badly wrong: "Whisper Small (~500MB)" is really about 563 MB on a graphics card, and closer to 928 MB without one. Sizes now reflect which of the two your machine will use.

• Settings → Storage now shows the real size of each downloaded model and where it lives on disk, and deleting one frees the space immediately.

• If you're upgrading, the old hidden model cache is cleared for you on first launch, which frees up the space it was holding.

• On macOS, fixed the download bar going missing when you closed the window and reopened Vocal Slice from the dock: the next model download reported nothing at all, so a fresh install looked frozen while it was quietly working.

• Fixed the Third-party licenses button in Settings → About, which did nothing at all in installed builds. The licences now open in a window inside the app, so you don't need anything installed to read them.

Which file?

• Windows — VocalSlice-Setup.exe
• macOS — VocalSlice.dmg, then drag Vocal Slice to Applications. VocalSlice-macOS-update.zip is the package the app updates itself from — not a manual download.

SHA-256 checksums

File	Size	SHA-256
VocalSlice-macOS-update.zip	198.5 MB	7e180ed9457c61d6b1a564f96b1aa81b1c824e0b7d2c89d8d4002c9e06fed0f7
VocalSlice-Setup.exe	94.4 MB	28ce0fbc726d65f221f97419183011f7ecc10a79c5aba959b1473aefae4cc208
VocalSlice.dmg	205.5 MB	12e0122eb6ea34e6292535977ee735dda7a29b59da833ab3af29aed8863eef96

Verify: certutil -hashfile <file> SHA256 (Windows) · shasum -a 256 <file> (macOS)

---

Windows isn't code-signed yet, so SmartScreen warns on first run — choose More info → Run anyway; the checksum above is how you verify the download. macOS is signed and notarized.

1.462.0

What's new

• Settings → About now lists what changed in each release, newest first, so you can catch up on what you've been given without leaving the app. It's read from the copy shipped inside the app — nothing is fetched.
• When an update has installed, Vocal Slice now says so once and offers to show you what changed. Previously it updated silently and you had no way of knowing what was new.
• The first-run tour is worded more plainly.
• Your remaining trial days are now shown at the top of the window, next to the theme and GPU icons. Previously the countdown was only in Settings → License, so a trial could run out without you ever seeing a number. It turns amber for the last two days, and clicking it takes you to your licence options.
• When the trial ends, Vocal Slice now explains what's happened and offers to buy a licence or enter a key, instead of a bare message that dropped you into Settings.
• Creating a slice after the trial has ended now tells you why it wasn't saved. It used to fail quietly and the file was lost when you closed the app.

Which file?

• Windows — VocalSlice-Setup.exe
• macOS — VocalSlice.dmg, then drag Vocal Slice to Applications. VocalSlice-macOS-update.zip is the package the app updates itself from — not a manual download.

SHA-256 checksums

File	Size	SHA-256
VocalSlice-macOS-update.zip	198.5 MB	8f7a41a916a8d417a73871d0c36b83198886b5e78d1ce45ce88ebd74048a1170
VocalSlice-Setup.exe	94.4 MB	de8f3cca9faf7a4a90405b3778030f9f431db0396ccea238b43188a5c7f0baa8
VocalSlice.dmg	205.5 MB	c156fd2a9477c6c73f1e13004fe926f4f8d3bdc17d221237430ad9ea10904d2d

Verify: certutil -hashfile <file> SHA256 (Windows) · shasum -a 256 <file> (macOS)

---

Windows isn't code-signed yet, so SmartScreen warns on first run — choose More info → Run anyway; the checksum above is how you verify the download. macOS is signed and notarized.

1.451.0

What's new

• Vocal Slice now transcribes in all 99 languages Whisper knows, instead of the 14 that were listed. Greek, Turkish, Polish, Hebrew, Swahili and the rest are all there. Slicing works exactly as it does in English — select the words you want and cut. Pick one of the multilingual models for this; the English-only ones still only do English.
• The language picker is now a search box: start typing and it filters, so you don't scroll a list of 99. It finds a language by its English name, its own name, or its code — "greek", "Ελληνικά" and "el" all get you there, and accents are optional, so "espanol" finds Spanish.
• Your chosen language is no longer forgotten when you switch between two English-only models.
• You can now tick several slices in the Slices tab and export them all at once — pick a folder and they're saved into it with your filename template, instead of exporting them one at a time.
• You can drag a slice straight out of the app onto your desktop or into your DAW. Grab it anywhere on the slice except the playback controls or the waveform. If you've ticked several slices, dragging any one of them drags the whole set.

Which file?

• Windows — VocalSlice-Setup.exe
• macOS — VocalSlice.dmg, then drag Vocal Slice to Applications. VocalSlice-macOS-update.zip is the package the app updates itself from — not a manual download.

SHA-256 checksums

File	Size	SHA-256
VocalSlice-macOS-update.zip	198.5 MB	db0e4a1bf9836e7834aab588ae74cce281da2f9601a8d1db3ce692b986612694
VocalSlice-Setup.exe	94.4 MB	2b60d058700e6ea9b38c78342e7d763fff707b28fc80c83d4b2917d0e2061c4a
VocalSlice.dmg	205.5 MB	90f880d4e0286a099a34fa40d54f0f6c48dec248b83c1d00caa522b20fa83210

Verify: certutil -hashfile <file> SHA256 (Windows) · shasum -a 256 <file> (macOS)

---

Windows isn't code-signed yet, so SmartScreen warns on first run — choose More info → Run anyway; the checksum above is how you verify the download. macOS is signed and notarized.

1.446.0

What's new

• Transcription progress now shows on the app's taskbar button on Windows and its dock icon on macOS, so you can leave a long file running and check on it without switching back to the app. When it finishes the icon flashes if you're in another app.
• Settings → License now tells you what happens when the trial ends, instead of listing what the trial includes.
• The Subscribe button in Settings → License no longer has its text underlined.
• The file picker no longer offers WMA and AIFF. Neither could actually be opened, and choosing one gave you an error saying transcription had failed.
• A file that can't be opened now says so, and names the formats that work, instead of reporting that transcription failed.

Which file?

• Windows — VocalSlice-Setup.exe
• macOS — VocalSlice.dmg, then drag Vocal Slice to Applications. VocalSlice-macOS-update.zip is the package the app updates itself from — not a manual download.

SHA-256 checksums

File	Size	SHA-256
VocalSlice-macOS-update.zip	198.5 MB	d634eaacfaf523537b1ed6993924dcddf85b75ab283f3f0d6a31bad215641ee5
VocalSlice-Setup.exe	94.4 MB	562935d85aecf442724838cd163746366593ad535640479f6947fd1ef863f26b
VocalSlice.dmg	205.5 MB	48cca909056e22157dc92ce56ab49668a306d2cd0e7e784f0cf38e0d98c3b61d

Verify: certutil -hashfile <file> SHA256 (Windows) · shasum -a 256 <file> (macOS)

---

Windows isn't code-signed yet, so SmartScreen warns on first run — choose More info → Run anyway; the checksum above is how you verify the download. macOS is signed and notarized.

Whisper models v1

Whisper ONNX model weights downloaded on demand by Vocal Slice.

These are the onnx-community/*_timestamped conversions of OpenAI's Whisper models,
redistributed unmodified under Apache-2.0. The app verifies every file against the
SHA-256 checksums in the manifest it ships with, so these assets are not a trust
boundary.

Not an application release: it carries no installer and no latest.yml, and is marked
as a prerelease so the auto-updater ignores it.