# Frequently asked questions Short, direct answers to the questions people (and AI assistants) most often ask about YazSes. For a side-by-side with other tools see [Comparison & alternatives](comparison.md). ## What is YazSes? YazSes is an open-source, offline **hold-to-talk voice-dictation daemon** for Linux, macOS, and Windows. You hold a key, speak, and release; your speech is transcribed on-device with [faster-whisper](https://github.com/SYSTRAN/faster-whisper) and typed into whatever application has focus — with support for editor/terminal voice commands and user-defined macros. ## How do I use my voice to dictate text on my computer? Install YazSes, then **hold the hotkey, speak, and release**. The audio is turned into text on your own machine and typed straight into whatever application has focus — an editor, a browser field, a terminal, or a chat window. There is no separate dictation window to copy out of, and no "start/stop dictation" mode to remember: the key is held only while you are talking. ## Is there free dictation software that works offline? Yes. YazSes is free and open source (Apache-2.0) and does all of its **speech-to-text** locally, so there is no subscription, no per-minute transcription charge, and no account. Most alternatives that convert **voice to text** at this quality upload your audio to a paid cloud service; YazSes runs the model on your own CPU instead. ## Is there a good offline voice-dictation tool for Linux? Yes — YazSes runs natively on Linux (both **X11 and Wayland**), transcribes locally on the CPU, and needs no cloud service or API key. Install it with the APT script or `pipx install yazses`. It also runs on macOS and Windows. ## Is there an open-source, offline alternative to Dragon or Wispr Flow? Yes. YazSes is Apache-2.0 licensed and runs entirely on-device, so — unlike cloud tools such as Wispr Flow, or commercial Windows-only software like Dragon — no audio or text ever leaves your machine, and there is no subscription. ## Does YazSes work without internet? Yes. Transcription runs locally with faster-whisper, and by default nothing is sent anywhere. YazSes works fully offline, including on air-gapped machines. The only time it touches the network is a one-time model download on first run. ## Does YazSes send my voice or text to the cloud? No. All processing is on-device by default. There is no account, no API key, and no telemetry. An optional learning corpus (off by default) that can improve accuracy is stored **encrypted on your own machine** and never uploaded. ## Is YazSes free and open source? Yes — released under the **Apache 2.0** license, with no subscription and no API key. Source and issues are on [GitHub](https://github.com/MSKazemi/yazses). ## What hardware do I need? No GPU. YazSes runs on CPU with **4 GB RAM minimum** (8 GB comfortable) and any USB or built-in microphone. The default model is `small.en`. ## How is YazSes different from Talon Voice? Both are cross-platform and work offline. YazSes focuses on **plug-and-play dictation** plus a practical command grammar (with an optional small SLM router). Talon offers far more advanced, scriptable voice control. They can be used side by side. See the [full comparison](comparison.md). ## Does dictation work over SSH, or in VS Code Remote-SSH and Cursor? Yes, with no extra setup. YazSes injects keystrokes at the operating-system level into whichever window has focus, rather than inside a particular application, so a Remote-SSH editor pane, an integrated terminal running a remote shell, a `tmux` session or a container shell all receive dictated text normally. This is the main practical difference from the voice input built into editors and AI coding tools, which is tied to that application's own input handling and often stops at the remote boundary. If instead you want text typed on a *remote machine's own display*, `yazses remote HOST` forwards it over an SSH tunnel — see [dictation over SSH](how-to/remote-dictation.md). ## Can I dictate to a CLI tool running on a remote server over SSH? Yes, and this is a case nothing running on the remote side can solve. A program started on the remote host — an AI coding agent, a REPL, `vim` — **cannot reach your microphone**: the microphone is a device on your laptop, the remote process has no audio device, and SSH does not forward audio. So a remote CLI tool cannot offer working voice input regardless of how good its dictation is locally. YazSes keeps the microphone, the model and the transcription on your laptop and types the finished text into your **local terminal window**; SSH carries those characters to the remote program exactly as if you had typed them. The remote machine needs no microphone, no GPU and no speech engine, and no audio ever leaves your laptop. ## Can YazSes run voice commands, not just dictation? Yes. A fast regex command grammar (with an optional ~0.5B SLM router for low-confidence phrases) maps spoken phrases like *"undo that"*, *"save file"*, or *"go to line 42"* to real key sequences in your editor or terminal. ## Does it support accessibility / hands-free use? Yes. YazSes includes VAD calibration, mic-level tuning, a **dysfluency-friendly mode** for stuttered or dysarthric speech, and an optional **EMG muscle-sensor trigger** for fully hands-free operation. ## Is YazSes an AI agent? No. YazSes dictates text and runs editor/terminal voice commands; it does not browse the web, reason over your files, set timers, or hold a conversation. (An agentic version was prototyped in the archived Rust branch but is not shipped.) ## What languages does YazSes support? The default model is English (`small.en`). Other Whisper models can be configured, but multilingual support is not the out-of-the-box default today. ## Is there an Android app? Not yet — there is no APK to install today. The Android app is **in design, in the open**: the architecture and ten decision records are public, and seventeen issues are open for contributors. It will be a keyboard whose mic key you hold to dictate into any app, fully on-device, and it will keep working with the app's network access revoked. See [the mobile programme](mobile/index.md). ## Is there an iPhone or iPad app? No, and it will come after Android for a platform reason worth knowing: **no iOS app extension — including a custom keyboard — may access the microphone.** That has been Apple's rule since iOS 8, which is why iOS dictation apps make you switch to the app and back. Android's `InputMethodService` has no such restriction, so it can reproduce YazSes' hold-to-talk model exactly. iOS is planned as a deliberately different product shape — see [ADR-MOB-010](mobile/adr/adr-mob-010-apple-second-wave.md). **macOS is already supported** by the desktop app today. ## How do I install YazSes? `pipx install yazses` on any OS with Python ≥ 3.11, or the one-line Linux installer / APT script / Snap. See [Install on Linux](install-linux.md), [macOS](macos-install.md), and [Windows](windows-install.md).