OFFICIAL FACT SHEET · SINGLE SOURCE OF TRUTH
Supersey Product Specifications, Privacy Boundaries & Licensing Facts
This page serves as the canonical reference for objective technical specifications, offline vs. network boundaries, hardware requirements, and one-time pricing across Supersey for macOS and Windows.
Last verified: 2026-11-14
1. Current Releases & Installer Specifications
macOS and Windows releases are versioned and packaged independently. Installers ship the runtime without bundling multi-gigabyte speech weights inside the initial download.
| Current macOS release | 0.19.0 (Supersey-0.19.0.dmg, ~238 MiB, SHA-256 served live by /api/releases/latest) |
|---|---|
| Current Windows release | 0.20.0 (Supersey-Windows-Setup-v0.20.0.exe, ~198 MiB, SHA-256 served live by /api/releases/latest) |
| What changed in Mac 0.17.0 | The voice command key can be changed in Settings; the browser extension's translation window is no longer covered by a page's own header, and the selection icon sits in a handier place. Includes voice commands and quick undo from 0.16.0. The native menu needs macOS 13 or later; earlier systems keep the previous menu. Upgrading from 0.14.3 or later keeps the granted permissions; re-export the browser extension from Settings and reload it. |
| What changed in 0.14.2 | Rotates the license signing key: 0.14.0 and earlier must upgrade (install 0.19.0 on Mac, 0.20.0 on Windows), and going online once after upgrading restores the license without re-activation. On Windows, recording with a Bluetooth headset no longer reconnects the microphone while the headset switches to hands-free mode; text and screen translation still open when no microphone or speech model is ready; URLs, IP addresses and file paths stay intact in formatting. On Mac, the text translation window frees memory when hidden. Upgrading from 0.14.0 needs no new Chrome extension export; from earlier versions, export it again from the app and reload it. |
| Default hold-to-talk hotkey | macOS: hold Right ⌘ (Command); Windows: hold Right Alt. Both can be rebound in Settings |
| Cursor text injection (Office & Vibe Coding) | Simulated keystrokes, clipboard paste, and chunked typewriter output — built for Vibe Coding long AI prompts in Cursor, VS Code, Claude Code, and terminals (automatically skipping newlines in terminal windows so Enter is never triggered early) as well as office drafting (stutter folding, long-passage paragraph numbering, and 'wrong=right' glossary replacement) |
| Bluetooth headset recognition optimization | Supports system microphones and Bluetooth headsets. Capture depends on the device, driver and environment; if the waveform stays flat, check the system input device and microphone permissions |
2. System Requirements & Local Weight Cache Size
On first launch, Supersey downloads only the local speech, translation, and speech-readout weights required for your enabled features into a local cache folder, then runs offline.
| macOS hardware & OS | Apple Silicon (M1 / M2 / M3 / M4), macOS 13 Ventura or later, accelerated via unified memory |
|---|---|
| Windows hardware & OS | Windows 10 / 11 (64-bit); supports NVIDIA discrete GPUs via CUDA (6 GB+ VRAM recommended) as well as CPU-only local execution |
| Recommended system memory | 8 GB RAM minimum (16 GB recommended when running voice dictation and local translation together) |
| Dubbing languages and downloads | Local voices support Chinese and English; other languages use voices installed in the operating system. The first download is about 330 MB on Mac or 379 MB on Windows. Macs with the Chinese model need only about 1 MB of English voices; existing Windows models need no extra download. |
| Local weight disk footprint | ~1.5 GB – 2.5 GB downloaded on demand into the user cache directory on first run; removable at any time |
3. Offline Execution vs. Network Activity Boundaries
Microphone audio and screen-capture regions are processed strictly in local memory and are never uploaded to cloud inference servers.
| 100% local / offline operations | Hold-to-talk voice dictation, spoken voice-to-foreign-language translation, screen region OCR & translation, and local spoken readout of translated subtitles |
|---|---|
| Network event A: First-run weight download | Connects only when fetching local speech/translation weights on initial setup or when enabling a new module |
| Network event B: License activation & lease renewal | Connects to license.supersey.org upon activation, unbinding, and once every 14 days to renew the lease (sends only SHA-256 hashes of the key and device fingerprint, with a 14-day offline grace period) |
| Network event C: Web pages & online video subtitles | The browser extension runs on pages you browse; when online video subtitle fetching is enabled in the desktop app, your machine requests public subtitle tracks directly from the video site |
4. One-Time Pricing, Editions & Device Allowance
Sold as a one-time purchase with no monthly or annual subscription and no word or minute quotas.
| Basic Edition — $15 USD (one-time) | Includes hold-to-talk voice dictation, spoken foreign-language translation, bilingual web page & selection translation, and bilingual video subtitles |
|---|---|
| Plus Edition — $25 USD (one-time) | Includes everything in Basic + synchronized spoken audio readout of translated video subtitles with automatic source audio ducking |
| Complete (Pro) Edition — $35 USD (one-time) | Includes everything in Plus + on-screen region capture OCR & translation for unselectable video frames, scanned PDFs, games, and remote desktops |
| Devices per license key | Up to 2 concurrent devices per key (macOS and Windows can be mixed); self-service unbinding is built into the app for moving to a new computer |
5. Supported Languages & Custom Glossary
Tuned for low-latency everyday dictation and bilingual workflows, including mixed Chinese-English speech without switching modes.
| UI & documentation locales (5) | Simplified Chinese (zh-CN), Traditional Chinese (zh-Hant), English (en), Japanese (ja), Korean (ko) |
|---|---|
| Speech recognition & translation coverage | Chinese (Mandarin, Cantonese, and mixed Chinese-English), English, Japanese, Korean, French, German, Spanish, Russian, and other major languages |
| Custom glossary & deterministic replacements | Supports custom vocabulary terms plus deterministic 'wrong=right' replacement rules and Simplified/Traditional Chinese conversion |
Reproducible Benchmarks & Comparison Guides
Each guide below includes side-by-side platform comparisons and a 5-step reproducible test protocol you can run on your own machine:
BENCHMARK GUIDE 01 · WINDOWS 10/11
Windows Local Voice Typing Benchmark
Windows 10/11 includes Win+H voice typing, which relies on cloud speech endpoints and frequently drops out offline, on restricted corporate networks, or during mixed-language technical dictation. Supersey for Windows (0.9.2) runs speech recognition and translation locally on your NVIDIA GPU (CUDA) or CPU — hold Right Alt to speak, release to insert text at your cursor.
Windows Local Voice Typing: Comparing Supersey 0.9.2 with Built-In Win+H →BENCHMARK GUIDE 02 · MACOS (APPLE SILICON)
Mac Local Voice Typing Benchmark
Apple Silicon Macs (M1 / M2 / M3 / M4) feature unified memory well suited for local speech and translation inference. Supersey for macOS (0.9.5) lets you hold Right ⌘ to dictate or translate directly into any cursor position, and pairs Apple's native Vision framework with local translation for on-screen OCR.
Mac Local Voice Typing: Apple Silicon Offline Dictation vs. macOS Built-In Dictation →BENCHMARK GUIDE 03 · PRICING & ARCHITECTURE
One-Time Purchase vs. Subscription
Most modern AI dictation tools charge $12–$15 per month ($144–$180 per year) because every spoken sentence is streamed to cloud GPUs. Supersey runs speech recognition and translation on your own Mac or Windows hardware, enabling a straightforward one-time purchase ($15 Basic / $25 Plus / $35 Complete) with zero recurring fees.
Local One-Time Purchase ($15–$35) vs. Cloud Subscription ($15/mo): TCO & Privacy Comparison →SUPERSEY · macOS & WINDOWS
Just say it.
The next message, the next note, the next page you cannot read.