Skip to content

user-guide.md

Latest commit

 

History

History
143 lines (90 loc) · 5.33 KB

user-guide.md

File metadata and controls

143 lines (90 loc) · 5.33 KB
Raw

TLDR User Guide

Overview

TLDR takes long text and produces a focused summary using a local AI model (Phi-4 Mini via Microsoft Foundry Local). Everything runs on your machine with no cloud dependency.

Input Methods

Paste from Clipboard

Press Ctrl+V at any time. If you're viewing a result or listening to a reading, the current operation stops and TLDR loads the new clipboard text.

Drag and Drop

Drag any supported file onto the window:

Format Extension
PDF .pdf
Word .docx
Plain text .txt
Markdown .md

Limits: 10 MB per file, 2,000,000 characters maximum.

Global Hotkey

Press Ctrl+Shift+S from any application. TLDR opens (or comes to the foreground), reads the clipboard, and loads the text automatically.

Summary Options

Style

Select one of six output formats before distilling:

Style Description
Bullets Markdown bullet list of key points
List Numbered list ordered by importance
Table Three-column table: Topic, Key Points, Details
Prose Flowing paragraph summary
Simple Plain everyday language, no jargon or technical terms
Same Condensed version preserving the original formatting

Detail Level

A slider with three positions controls how much of the original content is retained:

  • Brief: ~20% of original length, critical points only
  • Standard: ~35%, all main points covered
  • Detailed: ~50%, supporting details and examples preserved

Tone

Chip toggles in the main window, below the detail level:

  • Neutral: No tone modification (default)
  • Formal: Professional, academic language
  • Casual: Conversational, approachable language

Working with Results

Copy

Click Copy to place the summary on the clipboard in two formats:

  • Rich HTML: Pastes with formatting into Word, Outlook, Teams, etc.
  • Plain Markdown: Falls back to raw text in editors that don't support HTML paste.

Re-distill

Click Re-distill to return to the Loaded state with your original text intact. Change style, detail, or tone, then Distill again.

Read Aloud

Click Read Aloud to hear the summary spoken sentence by sentence. The current sentence highlights in the output view as it's read. The voice can be changed in Settings.

Controls during playback:

Button Action
Pause Pauses speech; click again to resume
Stop Stops speech and returns to the result view

Press Ctrl+Shift+X from any application to stop reading immediately.

Voice

Open Settings (gear icon) and select a voice from the dropdown. The list shows available English neural voices from the Edge speech service. Your selection is saved and restored on next launch.

The TTS engine falls back automatically: Edge neural voices (best quality, needs network) > WinRT voices > SAPI5 (offline, always available).

Theme

Open Settings (gear icon) and choose:

  • System: Follows Windows dark/light preference
  • Dark: Forces dark mode
  • Light: Forces light mode

The theme applies to both the app chrome and the summary output view.

System Tray

Closing the window minimizes TLDR to the system tray instead of exiting. Use the tray icon to:

  • Double-click: Restore the window
  • Right-click > Open: Restore the window
  • Right-click > Exit: Quit the application

Machine Info

The Settings popup includes a machine configuration box showing:

  • AI model name and context window size
  • CPU core count and available RAM
  • Active TTS engine

This helps verify which hardware resources the app is using.

Settings Persistence

Your last-used Style, Detail, Tone, Theme, and Voice choices are saved automatically to %APPDATA%\Tldr\settings.json and restored on next launch.

Configuration

Advanced settings live in appsettings.json beside the executable:

Setting Default Description
Llm.Model phi-4-mini Foundry Local model alias
Llm.MaxOutputTokens 1024 Maximum tokens in the summary
Hotkeys.OpenWindow Ctrl+Shift+S Global hotkey to open and paste
Hotkeys.StopReading Ctrl+Shift+X Global hotkey to stop TTS
Window.MinimizeToTray true Close button minimizes to tray

Limitations

  • File size: 10 MB per file maximum.
  • Input length: 2,000,000 characters maximum. The model's 128k token context window is the effective ceiling; inputs that exceed it (after reserving space for prompt and output) are rejected with a message.
  • Languages: Best results with English. Other languages may produce lower quality summaries.
  • First launch: Model download (~2.4 GB) requires an internet connection. After that, TLDR works fully offline.
  • AI disclosure: Every summary shows an AI-generated content footer. If the summary is longer than the input, a warning banner flags possible hallucinated content.