Skip to content

karandeepbhardwaj/Yapper

BranchesTags

Repository files navigation

Yapper

Voice-to-text desktop app that captures speech, refines transcripts with AI, and auto-pastes at your cursor

Build Status Version License Platform PRs Welcome

Onboarding Tutorial

macOS Windows
macOS Tutorial Windows Tutorial

Features

  • 100% local, private by design -- speech recognition and AI refinement both run on your machine. No cloud APIs, no API keys, no external apps. Your audio and text never leave the device.
  • Voice capture -- press a global hotkey and start talking
  • Voice commands -- speak translate, summarize, draft, explain, or chain to trigger AI actions directly; classified before refinement
  • Local speech-to-text (Whisper) -- on-device transcription via whisper.cpp; pick a model size (tiny → large-v3) in Settings, downloaded once
  • Local AI refinement (Ollama) -- transcripts are cleaned up by an open model (e.g. Llama 3.2) running in your local Ollama server; no API key in the app
  • Auto-paste refined text at your active cursor position
  • Conversation mode -- back-and-forth AI chat with a dedicated hotkey (Cmd+Shift+Y / Ctrl+Shift+Y), session summaries saved to history
  • Recording modes -- "Press" (toggle, default) or "Hold" (press-and-hold to record, release to stop; Fn key release supported on macOS)
  • Help screen -- "How to Yapp" in-app guide with voice command reference
  • Onboarding tutorial -- platform-specific animated tutorial (macOS dock / Windows taskbar) showing widget lifecycle, email paste workflow, and history dashboard
  • Dictionary -- user-defined text replacements applied before AI refinement (e.g., "btw" -> "by the way"), handles trailing punctuation
  • Snippets -- reusable text templates that bypass AI using word boundary matching (e.g., "my email" -> expands to your email address)
  • Style settings -- per-category refinement tone (Professional, Casual, Technical, Creative) for Email, Messages, Work, Personal
  • Code mode -- preserves code references in backtick formatting during refinement
  • Metrics -- usage tracking with streak days, word count, WPM, total recordings
  • Floating widget -- follows you across macOS Spaces, positioned above dock/taskbar (dock-aware on macOS, drops to bottom in full-screen), click-through when not hovered; shows error messages on failure
  • History dashboard -- fuzzy search (Fuse.js), pin/copy/delete with animations, sort by newest/oldest, multi-select category filter dropdown, action badges on cards
  • Theme persistence -- Light / Dark / Auto theme with circle-reveal transition animation
  • iOS-style transitions -- spring-based push/pop view transitions between app views
  • Settings page -- Whisper model picker, local AI (Ollama) model + server URL with live status, theme, hotkeys, recording mode, style, dictionary, snippets, metrics, code mode; segmented controls and hint tooltips. iOS 26 style "< Back" navigation
  • Customizable hotkeys -- dictation: Cmd+Shift+. (macOS) / Ctrl+Shift+. (Windows); conversation: Cmd+Shift+Y / Ctrl+Shift+Y
  • Fn key recording (macOS) -- use the Globe/Fn key as your trigger; Fn release stops recording in Hold mode
  • Atomic file writes -- all persistence uses write-to-tmp-then-rename to prevent data corruption

Architecture

Everything runs locally — no network calls leave your machine.

+----------------------------------+
|        Desktop App (Tauri)       |
|  +----------------------------+  |
|  |  React Frontend            |  |
|  |  (Tailwind + Motion)       |  |
|  +-------------+--------------+  |
|                | IPC             |
|  +-------------+--------------+  |
|  |  Rust Backend              |  |
|  |  - Global Hotkey           |  |
|  |  - cpal audio capture      |  |
|  |  - Voice Cmd Classifier    |  |
|  |  - Auto-paste / History    |  |
|  |                            |  |
|  |  whisper.cpp  ai_provider  |  |
|  +------+-----------+---------+  |
+---------|-----------|------------+
          |           | HTTP (localhost:11434)
   +------v-----+  +--v---------------------+
   | Whisper    |  | Ollama (local LLM)     |
   | model      |  | e.g. llama3.2          |
   | (on-device)|  | OpenAI-compatible API  |
   +------------+  +------------------------+

Installation

Desktop App

Download from the latest release:

Platform File
macOS (Apple Silicon) Yapper_x.x.x_aarch64.dmg
macOS (Intel) Yapper_x.x.x_x64.dmg
Windows (installer) Yapper_x.x.x_x64-setup.exe
Windows (MSI) Yapper_x.x.x_x64_en-US.msi

macOS Gatekeeper fix (unsigned app):

xattr -cr /Applications/Yapper.app

Windows permissions: Grant microphone access in Settings > Privacy > Microphone.

Required: local AI runtime (Ollama)

Yapper refines transcripts with a local LLM via Ollama. One-time setup:

  1. Install Ollama from ollama.com
  2. Pull a model: ollama pull llama3.2
  3. Make sure the server is running: ollama serve (the desktop app does the rest)

In Yapper's Settings → Local AI (Ollama) you can change the model name (default llama3.2) and server URL, and check live connection status. If Ollama isn't running, dictation still works — Yapper pastes the raw transcript and tells you AI refinement is unavailable.

Required: a Whisper model

On first launch, open Settings → Speech Recognition and download a Whisper model (start with base or small). Transcription runs fully on-device; nothing is uploaded.

Building from Source

Prerequisites

Dependency Version Install
Rust 1.75+ rustup.rs
Node.js 20+ nodejs.org
Bun latest bun.sh
CMake latest Required to build whisper.cpp (brew install cmake)
Xcode CLI Tools (macOS) latest xcode-select --install
Ollama latest ollama.com — local LLM runtime

Build & Run

git clone https://github.com/karandeepbhardwaj/Yapper.git
cd Yapper

bun install

# Development mode (hot reload)
bun tauri dev

# Production build
bun tauri build

Build output: apps/desktop/src-tauri/target/release/bundle/

How It Works

 Speak  -->  Record  -->  Transcribe  -->  Classify  -->  Refine/Execute  -->  Paste
  |            |              |               |                 |                 |
  |       Microphone     whisper.cpp     Voice cmd?        local Ollama       Keystroke
  |       (cpal)         (on-device)    (translate,       (llama3.2)         simulation
  |                                      summarize,       on localhost      (auto-paste)
  |                                       draft…)
  1. Speak -- press Cmd+Shift+. / Ctrl+Shift+. (or click the floating widget, or press Cmd+Shift+Y / Ctrl+Shift+Y for conversation mode)
  2. Record -- audio is captured from the microphone via cpal
  3. Transcribe -- whisper.cpp converts speech to text fully on-device
  4. Classify -- AI-first intent classifier detects voice commands (translate, summarize, draft, explain, chain) and dispatches them; non-commands proceed to refinement
  5. Refine -- the transcript is sent to your local Ollama model over localhost:11434
  6. Paste -- the refined or command-executed text is automatically pasted at your current cursor position

AI Model Configuration

Yapper refines text with a local model served by Ollama — no API keys, no cloud. Configure it in Settings → Local AI (Ollama).

  1. Install Ollama and pull a model: ollama pull llama3.2
  2. (Optional) Set a different model name or server URL in Settings. Any chat model in your Ollama library works — e.g. llama3.1, mistral, qwen2.5.
  3. Use Test model in Settings to confirm it responds.
Setting Default Notes
Model llama3.2 Any model pulled into Ollama
Server URL http://localhost:11434 Override with the YAPPER_OLLAMA_URL env var too

If Ollama isn't running, Yapper pastes the raw transcript unrefined and surfaces a "Local AI not running" notice.

Voice Commands

Once AI is configured, you can use voice commands by starting your recording with:

Command Example phrase Action
translate "translate this to French: ..." Translates the spoken content
summarize "summarize: ..." Produces a concise summary
draft "draft an email to the team about..." Generates a full draft
explain "explain what a closure is" Explains a concept
chain "translate then summarize: ..." Chains multiple actions

Configuration

Hotkeys

Function macOS Default Windows Default Customizable
Dictation Cmd+Shift+. Ctrl+Shift+. Yes -- in Settings
Conversation Cmd+Shift+Y Ctrl+Shift+Y Yes -- in Settings
Fn key Fn (Globe key) N/A macOS only

Recording Modes

Mode Behavior
Press (default) Press hotkey to start, press again to stop
Hold Hold hotkey to record, release to stop (Fn key release also stops on macOS)

Configurable in Settings.

Settings

Settings are persisted per-platform in the app config directory using atomic file writes:

  • macOS: ~/Library/Application Support/com.yapper.app/settings.json
  • Windows: %APPDATA%/com.yapper.app/settings.json
Setting Default Description
ollama_model llama3.2 Local LLM model name (must be pulled in Ollama)
ollama_url http://localhost:11434 Local Ollama server URL
whisper_model -- Downloaded Whisper model (tiny → large-v3)
whisper_language auto Transcription language, or auto-detect
theme Auto UI theme: "Light", "Dark", or "Auto"
hotkey Cmd+Shift+. / Ctrl+Shift+. Dictation hotkey
conversation_hotkey Cmd+Shift+Y / Ctrl+Shift+Y Conversation mode hotkey
recording_mode Press "Press" (toggle) or "Hold" (press-and-hold)
default_style Professional Default refinement style
style_overrides {} Per-category style overrides
metrics_enabled true Usage metrics tracking
code_mode false Code reference detection

Widget States

+----------+       +----------+       +----------+
|          |       |          |       |          |
|   IDLE   | ----> |LISTENING | ----> |PROCESSING|
|          |       |          |       |          |
|  (gray)  |       | (orange) |       | (gradient)|
+----------+       +----------+       +----------+
     ^                                      |
     +--------------------------------------+
                  done / error
State Appearance Meaning
Idle Thin gray pill, expands on hover Ready to record
Listening Orange with wave bars + stop/cancel Recording speech
Processing Animated hue gradient Refining through AI

Tech Stack

Layer Technology
Desktop framework Tauri 2 (Rust)
Frontend React 18, TypeScript, Tailwind CSS 4
Animations Motion (Framer Motion)
Speech-to-text whisper.cpp via whisper-rs (on-device, all platforms)
Audio capture cpal (cross-platform, 16 kHz mono)
AI refinement Local Ollama model over its OpenAI-compatible API
Search Fuse.js (fuzzy search)
macOS interop objc2 + objc2-app-kit + block2
Windows interop windows crate (Win32 + WinRT)
Build tooling Vite, esbuild, bun workspaces
CI/CD GitHub Actions (macOS + Windows builds)

Contributing

Contributions are welcome! Please read CONTRIBUTING.md for details on setting up the development environment, code style, and how to submit pull requests.

License

This project is licensed under the MIT License -- see the LICENSE file for details.

Copyright 2026 Yapper contributors.

About

Voice-to-text desktop app that captures speech, refines transcripts with AI, and auto-pastes at your cursor

karandeepbhardwaj.me/Yapper/

Topics

desktop-app windows macos rust productivity ai speech-to-text transcription dictation voice-to-text tauri whisper-alternative

Resources

Readme

License

MIT license

Contributing

Contributing

Security policy

Security policy
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases 13

v0.4.0 Latest
Mar 31, 2026
+ 12 releases

Packages

 
 
 

Contributors

Languages