# QuickDictate - full brief > A tiny Windows tray app that types your speech into any app. Use six bring-your-own-key > cloud providers or optional fully offline Cohere and Whisper models. No subscription, > no QuickDictate account, no analytics unless you opt in. Home: https://quickdictate.lunarwerx.com/ Source: https://github.com/LunarWerxs/QuickDictate Download: https://github.com/LunarWerxs/QuickDictate/releases/latest License: PolyForm Noncommercial 1.0.0 (free for noncommercial use, source available; commercial use by separate license), https://github.com/LunarWerxs/QuickDictate/blob/main/LICENSE Pricing: https://quickdictate.lunarwerx.com/pricing.md Maker: LunarWerx Studios, https://lunarwerx.com/ Last updated: 2026-08-23 ## What it is QuickDictate is a Windows 10/11 (64-bit) tray app for voice dictation. Hold or tap a global hotkey, speak, and the transcript types straight into whatever window has focus: an editor, a chat box, a terminal, or any web form. There is no QuickDictate transcription server in the middle: cloud audio goes directly to the provider you picked using your own API key, and Local audio never leaves the PC. ## Who it's for Anyone who wants to talk instead of type on Windows: writers, developers, and anyone doing repetitive typing, plus anyone who wants speech recognition that never touches a QuickDictate-run server, either because they'd rather pick their own cloud provider or because they want the option to run fully offline. ## Providers (7 engines total) - ElevenLabs: cloud, streaming, bring your own key - Deepgram: cloud, streaming, bring your own key - OpenAI: cloud, streaming, bring your own key - AssemblyAI: cloud, streaming, bring your own key - DashScope: cloud, streaming, bring your own key - Google Cloud: cloud, batch (finishes after you release the hotkey), bring your own key - Local: fully offline, no key, no account, no internet connection needed after install Stack multiple API keys for one cloud provider and QuickDictate rotates through them, routing around any key that is dead or throttled. ## Local (offline) models Two options, managed entirely from Settings (install, select, cancel a download, delete): - Cohere Transcribe Q5: 1.65 GiB, the accuracy-first default - Whisper Large v3 Turbo Q5: 591 MiB, smaller, with broader language coverage Model weights download on demand into Local AppData rather than shipping inside the installer. Downloads run in parallel where supported and are verified before use. The selected model prewarms in the background and stays ready; switching back to a cloud provider releases its RAM and VRAM. ## Every feature - Cloud or fully local speech-to-text, switchable at any time - Types directly into whichever app currently has focus - Hold-to-talk or tap-to-toggle hotkeys, default F13 (hold) and F14 (toggle), both user-changeable to any key - "Scratch that" voice correction: say it and the last bit of dictated text deletes itself - A custom fix-list / text-replacement table for words it commonly mishears - Per-app dictation profiles that switch automatically based on which app is focused - Multiple API keys per provider, load-balanced with automatic failover - Settings stored in one plain, human-readable settings.json file: back it up, put it in git, or copy it to a new machine - Streaming feedback from five cloud engines; clear "processing" feedback from Google Cloud and Local, which finish after the hotkey is released - v0.5.0 reliability work: bounded buffers, chunked long Google Cloud recordings, less idle polling, and fixes for duplicate audio, emoji pasting, cold local results, and a Save & Restart bug ## Install 1. Download the latest installer from https://github.com/LunarWerxs/QuickDictate/releases/latest 2. Run it on Windows 10 or 11 (64-bit) 3. In Settings, either paste an API key for one of the six cloud providers, or switch to Local and install Cohere Transcribe Q5 or Whisper Large v3 Turbo Q5, no key needed 4. Hold F13 (or tap F14) and start dictating into any window ## Pricing QuickDictate is free for personal and any other noncommercial use under the PolyForm Noncommercial License 1.0.0, with the source on GitHub; commercial use needs a separate license from LunarWerx Studios (releases through v0.8.0 were MIT). No subscription, no QuickDictate account, no paid tier. Local mode costs nothing beyond the one-time model download. Cloud providers are bring-your-own-key: whatever ElevenLabs, Deepgram, OpenAI, AssemblyAI, DashScope, or Google Cloud charge for their own API is billed to you directly by that provider, not by QuickDictate. Full machine-readable breakdown: https://quickdictate.lunarwerx.com/pricing.md ## Privacy No sign-up, no dashboard, no QuickDictate account, and no analytics unless you opt in. With Local selected, audio is processed entirely on the PC. With a cloud provider, audio goes directly to only the service you chose, using your own key, and QuickDictate never sees or relays it. An optional daily GitHub check can install updates automatically, only after the download is verified. ## How it compares - Windows Voice Access / Voice Typing: built into Windows 11, free, and Voice Access dictates fully offline, but neither lets you choose a cloud engine, stack API keys, or set per-app profiles the way QuickDictate does. - Talon Voice: free, donation-supported, cross-platform (Windows/macOS/Linux) voice control and coding tool; focused on hands-free system control rather than a choice of cloud speech-to-text engines. - Dragon Professional (Nuance): paid, licensed, Windows-only software with pricing available on request rather than published; QuickDictate is free and provider-agnostic. - Wispr Flow: cross-platform (Mac/Windows/iOS/Android) freemium AI dictation app, 2,000 free words/week on desktop (1,000 on iPhone, unlimited on Android) then a paid subscription; QuickDictate has no word cap and no subscription of its own. ## FAQ Is QuickDictate free? Yes for personal and other noncommercial use (PolyForm Noncommercial, source available), no subscription, no account; commercial use is by separate license. Does QuickDictate work offline? Yes, via Local mode with the Cohere or Whisper model. Which providers can I use? ElevenLabs, Deepgram, OpenAI, AssemblyAI, DashScope, Google Cloud, or Local. What OS does it support? Windows 10 and Windows 11, 64-bit only. Does it collect data or require an account? No account, and nothing you dictate is collected. The app makes one call of its own: a once-a-day update check carrying a random install id, which can be turned off. An anonymized usage rollup (totals and provider mix, no text) is off unless you opt in. Can I use multiple keys per provider? Yes, with automatic rotation and failover. ## Limitations - Windows only: no macOS, Linux, or mobile build - Cloud providers require your own paid or free-tier API key from that provider; costs (if any) are the provider's, not QuickDictate's - Local models require a one-time download of 591 MiB to 1.65 GiB and disk space to match - No built-in cloud sync of settings across machines: settings.json is a local file you copy or back up yourself See the top of this file for links (home, source, releases, license, pricing, maker).