Talk to write, fix, and send.
Give your hands a break.

AI voice keyboard. Live. Hands-free.

Download for MacSee how it works
On Windows? Join the waitlist
6.9 thinking

Your voice does all the work, live and hands-free.

01Live streaming

Watch it type while you talk.

Each word shows up the moment you say it, right in your text field, instead of pasting in a whole block after you stop.

soink
Auto
Other dictation
Auto
02Auto Polish

Cleans up filler and grammar, live.

Drops the ums, false starts, and rambling, so what you say out loud comes out reading like you typed it.

Auto
03Voice editing

Fix a word just by saying it.

Say “change Tuesday to Wednesday” and it's done, no selecting and no retyping.

Auto
04Voice Send

Say “send.” It's gone.

When you're done, just say “send” and it presses enter for you.

Auto
05Voice + keyboard

Talk and type in the same sentence.

Move fluidly between voice and keyboard. Type a bit, talk a bit, with no shortcut to press again and nothing to restart.

Auto

If it has a text field, soink types there.

SlackGmailGoogle DocsChromeCursorTerminalChatGPTClaudeClaude CodePerplexityCodexGeminiNotionLinearFigmaVS CodeGitHubGitHub CopilotSlackGmailGoogle DocsChromeCursorTerminalChatGPTClaudeClaude CodePerplexityCodexGeminiNotionLinearFigmaVS CodeGitHubGitHub Copilot
MessagesWhatsAppTelegramDiscordSignalTeamsOutlookZoomMeetSuperhumanObsidianBearArcSafariXLinkedInMessagesWhatsAppTelegramDiscordSignalTeamsOutlookZoomMeetSuperhumanObsidianBearArcSafariXLinkedIn

Why I built soink.

Dear friend,

After years of heavy keyboard use, I injured my wrist and was later diagnosed with TFCC. To keep working, I started using my voice, and I tried every dictation app I could find.

They were all smart, but none of them worked the way I needed. Apple’s built-in dictation showed my words as I spoke, but it wasn’t accurate or smart enough to trust. The AI apps were the opposite: genuinely clever, but they’d have me talk, wait, then paste a block of text back. I couldn’t see my words as they formed, so I kept losing my train of thought, and I was still stuck on the keyboard afterward, cleaning up what they pasted.

So I built soink: a way to talk instead of type and keep my hands off the keyboard as much as I can. I made it to rest my wrists and keep working, and I hope it does the same for you, especially if your hands hurt the way mine did.

Mera HuangFounder of soink
Read the whole story →

Questions, answered.

soink is an AI voice keyboard for people who work at a keyboard all day and want to give their hands a break, and for anyone whose hands hurt, whether from wrist pain like TFCC, repetitive strain injury (RSI), or a disability.

It comes down to two things. It’s live streaming, so your words appear in the text field the moment you speak, cleaned up by AI as you go, and you see your text take shape in real time with low latency. And it’s hands-free: tap once to start, then keep talking, fixing a word or sending a message by voice, and dictating as long as you like.

When voice isn’t the right fit, you move fluidly between voice and keyboard, talking and typing in the same sentence.

It means writing, fixing, and sending all happen by voice, in one unbroken flow. Here’s how:

Tap once to start. No key to hold while you speak, and no second tap to stop. Talk for as long as you like, and when you go quiet, soink stops on its own.

Voice editing. Changed your mind mid-sentence? Just say the fix out loud: “change Tuesday to Wednesday,” and it’s done, with nothing to select or retype.

Voice Send. Done with your message? Just say “send,” and soink presses enter for you.

Continuous dictation. Sending doesn’t end the session. soink keeps listening after your message goes, so in a chat or an AI chatbot you can fire off one message, then the next, then the next, all by voice, without ever tapping to start again.

Put together, from your first word to sent and on to the next, your hands barely touch the keyboard.

Most dictation tools work the same way: record, transcribe, then paste a block of text once you stop. soink is different in a few key ways:

  • Live, right in your text field. No floating window, no paste step. Each word lands where your cursor is the moment you say it, instead of arriving as a block after you stop.
  • Hands-free, not just dictation. Most tools stop at writing, so you still reach for the keyboard to fix and send. soink keeps it all on your voice: fix a word by saying it, send the message with a “send,” then keep talking, so you can fire off one message and start the next without tapping again.
  • Voice and keyboard as one. Voice isn’t the right fit for every moment, and it doesn’t have to be. soink is a keyboard, so the keys are always there: type a bit, talk a bit in the same sentence, with no mode to switch and nothing to restart.

soink currently supports 11 languages across 16 variations: English (US, UK, India), Spanish (Mexico, Spain), Chinese (Mandarin and dialects), Arabic (Bahrain, Saudi Arabia), Portuguese (Brazil), French, German, Hindi, Italian, Japanese, and Korean.

More are on the way. Don’t see yours yet, or not sure if it’s covered? Email ask@soink.ai or ask in our Discord, and we’ll prioritize the languages people ask for most.

soink doesn’t log, save, or upload your typing or voice. Your audio becomes text and is then deleted, and we never train on your data. See our Privacy Policy for the details.

No, soink runs in the cloud. We chose cloud so we can give you the best speech recognition and AI models, and so soink runs on any Mac. On-device models are heavy, and not every computer can run them well, so the cloud keeps soink fast and accurate for everyone. We're actively working on local-first models, and as soon as one is good enough, we'll add an offline option too.

Give your hands a break.

Download for Mac
macOS 13.5+ · Apple silicon & Intel