Popfilter.ai
Coming soon

System-wide voice · on-device

Speak.Therestwritesitself.

Popfilter is a system-wide voice layer with local-first refinement and a hands-free agent loop, running entirely on your machine. No cloud. No signup.

Coming soonCross-platform
Welcome back
1 today|11 words
Your shortcuts
Push-to-Talk
Hold to record
Fn
Toggle Recording
Press to start/stop
FnShift⇧
Recent history
View all
June 10, 2026
nowDraft a reply to the review thread and summarize the changes
4:27 PMOk, what's pending for this particular work?
4:12 PMAlso, your tests don't really make sense. The prompts should be editable by the user.
4:12 PMRemove that spoken cap. I don't think we need it.
4:11 PMOkay, something does need to change. I like the gap.
All processing happens locally on your device
Speak anywhere✺Refine on-device✺Hands-free voice loop✺Your models✺Private by default✺No cloud✺No signup✺Speak anywhere✺Refine on-device✺Hands-free voice loop✺Your models✺Private by default✺No cloud✺No signup✺

What Popfilter does

Talk into any app

Hold to talk or tap to toggle, and your words land cleanly in whatever is in front of you. It notices the app automatically: casual in chat, precise in code, clean in email.

SSlack
casual

sounds good, let us ship it today

Adapts style per app

Speak. It answers.

A hands-free loop: you talk, the agent works, the reply is read back. One key.

Voice loopOne key, continuous

You speak

Hold one key and talk

Agent responds

Refined, then spoken back

You steer

Correct, redirect, repeat

12+ voices
ENDEESFRIT
1.0x

It finishes the thought

Type-ahead pulls referential context from your thread and files, then completes a full sentence in real time. Accept it and the ghost text becomes ink.

Did you get a look at the Q3 report?

Thanks Kate, I've attached the revised Q3 forecast.

Message Kate…

To-do · Today
Send James the Q3 forecast
Book the design review
Reply to the launch thread
Skim the analytics digest

Prep the Q3 review deck for Thursday

Your machine. Your models.

Choose the speech and language setup: built-in, your own local runtime, or Apple's on-device model, and wire it straight into the tools where the work happens.

Refinement

Choose your engine

Quick fill
Server URL
http://localhost:11434/v1

An OpenAI-compatible endpoint. Point it at a local server or your own key.

A glanceable cue

A tiny floating indicator shows when it is listening, refining, or has a reply ready. Drag it anywhere, or hide it.

ListeningProcessingReply ready
Your keys⌃Spacetalk⌃⇧Rtoggle

By the numbers

0
words / min
0+
apps
0+
voices
0%
on-device

On-device & yours

It all stays on your machine.

Your audio and text never leave your device. No cloud, no servers. Keep a searchable history of what you have dictated, or turn it off and set how long anything sticks around.

100% local · 0 sent
No cloudNo accountLocal-onlyApple Foundation ModelsYour own LLMs

Make it yours

Bend it to how you work.

Set your own shortcuts and drag the live indicator anywhere, or hide it. It launches at login and waits quietly until you need it.

drag anywhere · bottom · left
Per-app stylesCustom dictionarySnippetsPush-to-talk hotkeysSearchable history12 languagesPlace the indicatorMenu-bar controls