Frequently asked questions¶

Q: 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 and typed into whatever application has focus — with support for editor/terminal voice commands and user-defined macros.

Q: 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. New here? yazses quickstart prints the next steps for your machine and changes nothing.

Q: 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, both in preview: supported and installable, not yet verified on hardware by a person on a current build (see the validation matrix).

Q: 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.

Q: 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 base.en (141 MB).

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.

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 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. New here? yazses quickstart prints the next steps for your machine and changes nothing.

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, both in preview: supported and installable, not yet verified on hardware by a person on a current build (see the validation matrix).

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. By default the only network use is the one-time model download. Actions you start yourself — enabling a feature that needs extra packages or models, yazses update, Check for updates — also download; none of them uploads anything you said.

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.

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 base.en (141 MB).

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. Talon offers far more advanced, scriptable voice control. They can be used side by side. See the full comparison.

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.

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 I control YazSes with a brain-sensing headset?¶

Not by thinking words. Consumer EEG headsets do not read intended words, and YazSes does not try to. The experimental biosignal switch can use a muscle squeeze, a deliberate double blink or a jaw clench. These are muscle and eye signals that reach the headset's electrodes. The feature is experimental and off by default, and it has not yet been tested on real hardware. See hands-free computing with a muscle, blink or jaw-clench switch and muscle and brain control: what works.

Can YazSes run voice commands, not just dictation?¶

Yes. A fast regex command grammar maps spoken phrases like "undo that", "save file", or "go to line 42" to real key sequences in your editor or terminal.

Dictation is heard but nothing is typed on Ubuntu — what do I do?¶

Turn off the voice-activity overlay: yazses features disable overlay, then yazses restart. On Wayland the overlay window can take keyboard focus and the text is typed into it instead of your app. It is optional and off by default in current releases. The full explanation is in Troubleshooting.

Does it support accessibility / hands-free use?¶

Yes. YazSes includes VAD calibration, mic-level tuning, a dysfluency-friendly mode for stuttered or dysarthric speech. An optional, experimental EMG muscle-sensor trigger also exists as a hands-free alternative to holding a key; it is off by default and has no real-device evidence yet (see the known limits).

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, or hold a conversation. It sets timers only if you turn on the optional Local Voice Timer (yazses features enable voicetimer), which is off by default. (An agentic version was prototyped in the archived Rust branch but is not shipped.)

What languages does YazSes support?¶

The default model is English (base.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 being built in the open: the M0 foundations (Gradle skeleton, privacy gates, and Kotlin ports of the post-processing, command, vocabulary, audio and silence-gate cores) are in android/, and the first M1 work has landed, but nothing dictates yet. 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 for what is built and what remains.

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.

macOS already has a desktop app today, in preview — supported and installable, not yet verified on hardware by a person on a current build (see the validation matrix).

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, macOS, and Windows.