Kusema · Free for Windows
Talk to AI.
Keep your voice.
Private dictation and an AI assistant for law firms, professionals and students. Your voice never leaves your computer.
Windows 10/11 (64-bit). Early software, updated often. The installer isn't code-signed yet, so Windows may show “Windows protected your PC”. Choose More info, then Run anyway.
See it in 60 seconds
Demo video coming soon
For law firms
Small and solo firms
Paralegals and assistants too.
- Dictate letters, file notes and emails — text appears wherever your cursor is.
- Privileged mode keeps everything on your computer. No text leaves the building.
- Legal vocabulary and citation formatting built in: “42 U.S.C.”, “Smith v. Jones”, Latin terms.
- A familiar way to dictate if you're coming from Dragon: continuous dictation, spoken punctuation, “scratch that”, a Numpad + preset, and Dragon word-list import. Nuance has been winding down parts of its lineup — Dragon Anywhere is ending in 2026.
- Neulinc can set it up across the whole firm.
For professionals
Consultants, advisors & more
Anyone who writes a lot and handles confidential information.
- Write emails and reports by voice in any app — no copy-paste, no dictation app that only works in one place.
- Edit by voice: “make this more formal”, “make this shorter”, “translate to Spanish”.
- Set a style per app: email tone in Outlook, brief in Teams, formal in Word.
- Your voice stays on your computer.
For students
Free, no subscription
Works offline with a local AI.
- Dictate essays and notes by voice.
- Tutor style explains step by step instead of just giving you the answer.
- Ask about a lecture slide or textbook page on your screen — a preview shows what will be sent, and nothing is sent until you click Send.
- Use Ollama (free, local AI) and nothing needs a cloud account or subscription.
Follow your school's or exam's rules on AI use.
How it works
Your voice stays on your computer.
Your voice
You speak.
Transcribed on your PC
Speech becomes text locally. No audio is uploaded.
Only text goes to the AI you choose
Claude, OpenAI, OpenRouter, Groq and others. Or use Ollama or LM Studio to run AI on your own computer, and nothing leaves at all.
Reply spoken on your PC
The answer is read aloud by a voice on your computer.
Ask about my screen captures only the active window and shows a preview of exactly what will be sent and where. Nothing is sent until you click Send. Screenshots are kept in memory only and never saved.
Your API keys are stored in Windows Credential Manager, not in a plain-text file.
Controls
One key. Just talk.
Hold the key and talk: words are typed wherever your cursor is.
Double-tap for continuous dictation — text appears as you pause, through a whole letter or file note. Double-tap again to stop.
While holding, tap Space to ask Kusema by voice — or tap T to type your question. Or start with "Kusema, …"
Select text and ask Kusema to edit it ("make this more formal") — a preview appears before anything changes.
Ask about the window in front of you. A preview shows what will be sent. Nothing is sent until you click Send.
Cancels, or stops Kusema talking.
Right Ctrl is the Windows default; you can also use the Copilot key, Numpad +, or any key you pick. Updates install automatically.
Features
Built for the way you already work.
Continuous dictation with voice commands
Say “new paragraph”, “new line”, “scratch that”, or “stop dictation”. Speak punctuation out loud: “comma”, “period”, “question mark”, section sign, paragraph sign.
Voice editing and navigation
“Make this shorter.” “Fix the grammar.” “Translate to Spanish.” “Select last sentence.” “Delete that.” “Go to end of document.”
Vocabulary that learns
Includes a legal vocabulary pack and an optional medical pack. Add your own words and say “correct that” to teach Kusema a term once. What you teach it stays on your computer.
Privacy page
Shows every request Kusema sent off the computer: to whom, what kind, how large. Never the content. The header reads “Audio sent off this computer: 0 bytes.”
Privileged mode
One switch that blocks all cloud AI. In Privileged mode, Kusema uses only AI running on your own computer.
Local AI built in
Ollama and LM Studio appear in settings as “on this computer.” A fully private setup needs no cloud account.
Type-to-ask and paste the answer
Type instead of speaking. Add “…and type it” and the answer lands directly in your document: “Draft a polite reply declining this meeting and type it.”
Writing styles per app
Set a style for each application — formal in Word, brief in Outlook, concise in Teams and Slack. Applied when AI cleanup is on.
You always know when it’s listening
A glowing pill shows the recording state, with an optional screen-edge glow. If you switch windows mid-dictation, Kusema pauses.
Easy to set up
About two minutes from install to first dictation. 26 interface languages. Updates install automatically.
Privacy
Your voice is personal data.
A voice recording can identify you, and for lawyers, clinicians, accountants and advisors it often carries client information. Kusema keeps your audio on your device — no recording goes to a cloud service. The text of a question goes only to the AI provider you choose, under that provider's terms. Use Privileged mode to block all cloud AI, or check the Privacy page to see exactly what was sent. For complete privacy, run a local model through Ollama or LM Studio and nothing leaves your computer.
Audio never leaves your device
You choose where text goes
Fully local option
For organizations
Rolling Kusema out across your firm?
Neulinc helps law firms evaluate and deploy Kusema across their offices, including a fully private, in-office setup.
- Setup on every computer in the firm
- Privileged mode and policy settings — lock Kusema to local-only AI for your whole team
- A shared firm vocabulary and templates
- An in-office private AI server so nothing leaves the building (Neulinc private AI deployment)
- Offline installs for offices that restrict downloads
- Documentation your IT provider and compliance team can review
- Training and support
Neulinc also works with healthcare practices (therapists and small clinics), CPAs and financial advisors.
Book a callOpen Source
Open source, so you can check.
Kusema is free and open source under the MIT license, so its privacy claims can be verified in the code. View on GitHub →
Coming next
What’s planned.
Planned, not yet shipped.
Live preview of words as you speak
A microphone quality check
Legal and clinical templates: “file note”, “SOAP note”
“Ramble, then structure” — dictate freely and Kusema turns it into a structured document
Foot pedal support
Practice-management integrations: Clio; MyCase (for firms on the Advanced plan)
Kusema is built on Handy by CJ Pais, used under the MIT license. Kusema is independent and not affiliated with or endorsed by Handy.
FAQ
Common questions.
Yes. Kusema supports continuous dictation, spoken punctuation, voice commands (“new paragraph”, “scratch that”), Dragon word-list import, and a Numpad + preset. Many lawyers dictate with Dragon, and Nuance has been winding down parts of its lineup — Dragon Anywhere is ending in 2026.
Dictation always works offline. The assistant can too — with a local AI like Ollama, nothing needs a connection.
Any application with a text box: Word, Outlook, OnlyOffice, browsers and practice-management web apps. Voice editing works best where text can be selected.
ABA Formal Opinion 512 reminds lawyers that confidentiality duties apply when using AI tools. Kusema keeps audio on the device; Privileged mode keeps everything local; and the Privacy page shows exactly what was sent. Kusema is designed to help lawyers meet their own obligations — it isn’t certified for, approved under, or a substitute for the firm’s own review.
Yes, it is free for everyone. Tutor style explains step by step instead of just giving you the answer. Follow your school’s or exam’s rules on AI use.
Yes. Kusema is free and open source — for law firms, professionals and students alike. No subscription needed.
No. Speech is transcribed on your computer, and spoken replies are generated on your computer too.
The text of your questions goes to the AI provider you choose. A screenshot is sent only after you review the preview and click Send. Use Privileged mode to block all cloud AI, or check the Privacy page to see the full record. With a local model, nothing leaves your computer.
Anthropic (Claude), OpenAI, OpenRouter, Groq and others, or AI running on your own computer through Ollama or LM Studio.
Not yet. Kusema runs on Windows 10 and 11 (64-bit).
The installer isn't code-signed yet, so Windows shows 'Windows protected your PC'. Choose More info, then Run anyway.
Kusema isn't certified for any compliance framework. Your audio never leaves your device, and a fully local setup is available. Neulinc can help your organization evaluate Kusema within its own requirements.
That depends on the provider. Check your chosen provider's terms, or use a local model for full privacy.