Samsara is a local-first voice assistant for Windows. One persistent HANDS FREE lane: dictate naturally across pauses, fire exact commands between thoughts, say “finish” to paste. Your voice never leaves your machine unless you deliberately enable a cloud feature. Built first for one developer whose hands hurt; released for everyone’s.
“I'm Morne, and my hands fucking hurt.”
Samsara is an attempt to combine all of modern day's most powerful and effective accessibility systems into a single place. A place that is forever free, open source, and which doesn't require accounts or internet connections.
It's a goal: to take some of the weight off the shoulders of those in our society who are so often forced to carry heavier loads, with less support.
We need more accessibility advocates. We need better access to advanced accessibility systems. If any of these ideas resonate with you, please consider becoming a part of it by joining the beta. This project can't get funding without users. And no matter who you are, there are at least some features of this free application that can help you in your daily life.
— Morne
Speech becomes buffered dictation; exact commands — “scroll down”, “show numbers”, “switch window” — fire instantly without leaving the lane. “finish” pastes the whole thought. Hold-to-dictate and wake words still work too.
Speak in whole thoughts. Pauses do not cut you off; "end" pastes the lot. When a word comes out wrong, the correction is also a phrase.
Move, snap, focus and close windows, and open the app you want, without reaching for a mouse.
Numbered overlays put a label on anything clickable. Magnifier, colour filters, cursor and text size are all one phrase away.
Frequent commands stay deterministic. Open-ended ones can go to an assistant you configure — local, or your own cloud key.
Tabs, the address bar and history in the browser; playback, volume and what is on screen everywhere else.
Timers, reminders, task lists, screen recording, macros, smart lights, a 3D printer, and the keys and clicks everything else is built from.
Every change must remove a step — less friction, less cognitive load, less cost per interaction.
Software assumes interaction is cheap. For chronic pain, ADHD, or limited energy, it isn’t — Samsara starts from that.
Living with ADHD, "work smart, not hard" stopped being a productivity slogan and became survival logic. I was already working hard just existing in a room.
No marketing trick replaces engineering. Here's what's actually under it.
WASAPI capture, faster-whisper transcription — GPU optional, CPU fine. Local by design; network features are opt-in and labeled.
Live partials float in an overlay; the final paste swaps drafts through a clean undo-cycle.
Direct ctypes calls into IAudioEndpointVolume for volume control (because pycaw proved unreliable). TCP M-code protocol for FlashForge printers. JSON-RPC for Hyperion LED strips. AHK for Electron apps that ignore Win32.
Drop a Python file in plugins/commands/, add @command — that’s a voice command. Ships with plugins for smart home, music, printers, macros, screen recording, reminders, Ava, and more.
Frequent commands stay deterministic — no AI, no latency, no misunderstanding. Open-ended requests can route to AI you configure: local LLMs or your own keys. You set the boundary.
For the technically curious. You don't need to know any of this to use Samsara.
v0.22.1: HANDS FREE is one persistent lane — pauses no longer paste fragments; “end” commits the whole thought while the session stays live.
Also new: DOM-aware Show Numbers, high-DPI fixes, voice-managed vocabulary, config backup, OpenRouter for cloud Ava, local-only Task Lists.
Next: stabilize with real users, then the ambient Ava workflow — direct agents and triage information without touching a keyboard.
Shipped and beta items are generated from CHANGELOG.md in the Samsara source; nothing is ticked before it is in a build you can download.
Every roadmap item is filtered through the same question: does this reduce a step somewhere?
No installer — just extract and run. The standard download is the verified CPU build. The separate CUDA runtime pack is optional; no NVIDIA GPU is required.
Previous stable: v0.22.1 (.zip)
Windows only for now. Beta testers welcome — bug reports are the only currency this project accepts.
Extract, run Samsara.exe, follow the wizard. Hold the dictation key to talk — or turn on HANDS FREE and say “finish” when a thought is ready to paste.
No purchases, subscriptions, or limits — ever.
No account, no email, no telemetry. Core dictation never touches a server of mine.
AGPL-3.0, source public today — if I vanish, anyone can carry it on. You can’t be stranded.
Dictation, commands, wake words, history: local. Only opt-in features touch the network — model download, TTS, cloud Ava, your own webhooks.
Free forever — nothing is ever locked behind payment. Optional cloud AI just needs your own key.
Licensed under AGPL-3.0 — free and open source, including for commercial use.
Yes — the matcher is tolerant. Quiet voices do best with a close mic.
Whisper handles most English accents well; persistent misses go in the custom dictionary.
Yes — wake words mean never reaching for a key. A headset helps.
Not intentionally — it doesn’t touch focus or screen-reader output, but edge cases are untested. Report anything that breaks.
No — the standard build is CPU. An NVIDIA GPU can accelerate via the optional CUDA pack.
AGPL-3.0, public today. Forkable forever — you won’t be stranded.
Samsara is early-stage. Things will break. Telling me when they do is the most useful thing you can do for this project.
Samsara is free and stays free. If it saves your hands and you want to send a coffee, I won't say no — but please don't feel you have to.