What's new

1.474.0 26 August 2026

• 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.

1.462.0 18 August 2026

• 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.

1.451.0 11 August 2026

• 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.

1.446.0 1 August 2026

• 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.

1.437.0 27 July 2026

• First-run tutorial typo.

1.435.0 26 July 2026

• Fixed clips in the Slices tab being able to play over each other — editing a slice and previewing another could sound at the same time. Only one clip plays at a time now.
• Opening the file picker from the status bar no longer clears the demo (or your loaded file) when you cancel without choosing a new file.
• The first-run tutorial now also covers choosing a file, the model/language settings, and local GPU processing.

1.432.0 26 July 2026

First public release.

• Transcribes audio on your own machine — no uploads, no account, no telemetry.
• Cut clips by selecting words in the transcript; the waveform follows the selection.
• Lossless byte-level WAV slicing; other formats decode to 24-bit WAV.
• English and multilingual Whisper models, GPU accelerated via WebGPU with automatic CPU fallback.
• Filename templates, five themes, and a searchable transcript.
• Windows and macOS.