ScreenKit

Windows desktop tool (ScreenKit.exe, window title ScreenKit): screenshot recognition, annotation, recording, speech, translation, a dictionary, and face, plus file transfer and screencast with the Android companion. This PC also has an HTTP API and a browser tools page.

Current version: 1.0.26 · Releases · Changelog · 中文

Screenshot and recognition

• Screenshot OCR: A hotkey or the menu selects a region and recognizes text and barcodes; the recognition engine can be switched, and recognition can overlay a translation.
• Screenshot annotate: Rectangle, ellipse, arrow, pen, and text; on finish it copies as an image, a file, or a path.
• Repeat last capture: A hotkey or the menu grabs the last region you selected again.
• Long screenshot: Click a scrollable window and it scrolls and stitches one long picture.
• Screenshot history: Browse saved pictures as thumbnails and pan, zoom, annotate, or copy them.
• Overlay text: Draws the recognized text over the picture to select from.
• Clipboard: Paste a picture and recognize, annotate, or save it.
• PDF workbench: Open a PDF, recognize it page by page, edit the text, and export a searchable PDF.

Recording

Capture → Screen record or GIF record: click a window, or drag a region.

• The red frame is drawn outside the selection; size and aspect can be set before start, and the control bar can be dragged and collapsed. The audio source can be changed while recording.
• Region recording: Writes an MP4, with system audio, the microphone, or both, and optional mouse and click highlights.
• GIF record: Uses the same region capture and writes a silent GIF, with a frame rate and a palette.
• Record options: The frame rate and the same shorter-side cap as screenshots.

Speech, translation, dictionary, and face

• Speech recognition: Live captions with an offline or streaming model.
• Speech synthesis: Windows speech, SAPI, offline speech, Edge natural voices online, and voices in the local Chrome. Chrome lists only installed languages, and the page opens Chrome model installation.
• Voice input: The hotkey starts and finishes it, and the result goes to the current input position.
• Chat: Uses a configured LLM, with bubbles, microphone input, and speaking.
• LLM log: The latest ~1,000 requests with the model, elapsed time, and the request and response.
• Translation: Local ONNX, a configured LLM, or the local Chrome. Auto mode translates between the native and learning languages and shows the detected direction; the tab and popup pick the engine and model. Chrome language lists show only installed packs.
• Face: Detection and comparison, with optional landmarks and gender and age.
• Chinese to pinyin: Full spelling without tone marks, or one lowercase initial per character.
• Wubi, Japanese, and Korean IME: Type with a candidate bar in this app and in other programs. Turn it on from Tools, and manage custom words in the tray Lexicon.
• Simplified/Traditional: Simplified on top, traditional below, following Taiwan character forms.
• Japanese yomi: Turns a Japanese sentence into readings, by word or by character.
• Calendars: Gregorian to lunar (sexagenary cycle and leap months), plus Japanese, Taiwan, Hebrew, Hijri, Persian, and others.

Dictionary

The main window, or Tools → Dictionary. Look up Chinese, English, Japanese, and Korean.

• Language filter: All, Chinese, Japanese, Korean, English.
• Pure Latin input is romanized for Japanese and Korean search (romaji to hiragana, Korean romanization to Hangul), such as watashi and hanguk.
• A hit has pronunciation, part of speech, senses, phrases, and examples, all of which can be spoken; Export writes every hit of this search.
• Double-click, drag, or right-click inside an entry to pop up the lookup window, up to 100 hits. Dictionary entries always do this.
• Settings → Dictionary sets how many hits to list, how long to keep the database open, the selection popup, and each language's engine, voice, and rate.
• The dictionary hotkey is empty by default: it brings the window forward on the dictionary when it is not the main window, and hides the main window when it is.
• The web dictionary is its own page at /dict, with the same search, four languages, speech, and export.
• Selection popup is off by default and can be turned on on the dictionary page, under Settings → Dictionary, or on the tray. While it is on, dragging or double-clicking text in this program or another shows a lookup popup that does not take focus. Drag the lower-right corner to resize; the size lasts until you quit.

Files and screencast

File sync

On the PC, open File sync; the phone and the PC are on the same LAN.

• The list is sendfile/ next to the program, one level deep: list or thumbnails, marquee and multi-select, cut/copy/paste, and delete to the Recycle Bin.
• Dropping files onto the list puts them in the inbox; dropping onto the lower area sends them to the phone while it is connected, and a phone share or upload writes to sendfile/.
• Web manager uses the same port: the PC opens /files and the phone /m, where a login allows uploads, folders, rename, and delete.

Screencast

This program receives while it is running, and that can be turned off in Settings; Tray → Tools → Screencast opens the screencast page. Use the same Wi-Fi, with guest isolation off.

• PC to PC: Pick a saved PC, scan the LAN, or type an IP; the picture is kept within 1000×1000 by default.
• Phone to PC: Pick a saved PC or scan the LAN in the phone app, over the LAN or a USB cable (debugging or accessory).
• The viewer is not pinned on top, the picture can fit the window or fill it, closing it disconnects, and quality and the audio source can be changed while casting.

Tools

• Image convert: Batch convert to JPG, PNG, palette PNG, or BMP, with a shorter-side cap, rotation, and mirror.
• Batch video convert: Batch convert to MP4, MKV, or WebM, with a shorter-side cap and the audio kept.
• Video to GIF: Pick a start and an end and write a silent GIF, with a frame rate and a palette.
• Character art: The Mode list picks Shades ASCII, Dot shading, or one of the line drawings; full-color character art is drawn in a console window.
• Pixel art: Choose a width and a height, paint on large cells, and preview at real pixel size. Copy or save a PNG.
• Character bitmaps: Draw each character as a dot matrix and export a C array, hex, or dot-matrix text.
• QR / barcode: Generate QR, Data Matrix, Code 128/39, EAN, and UPC.
• Batch rename: Edit target names by hand or with rules such as %1 and #, and sort by name or date.
• Fast copy: Copy, cut, delete to the Recycle Bin, or make a symbolic or hard link, and register an Explorer cascade menu.
• Hash: MD5, SHA-1, and SHA-256. Paste an expected value to compare.
• Text tools: Base64, URL, hex, escapes, case, whitespace, counts, JSON pretty-print, and more.
• Markdown reader: Open or drop a Markdown file and read it with the IE11 engine built into Windows.
• Password generator: Generate passwords from a length and character classes, or hand a sentence to a configured LLM and spell the translation out.
• Network: Diagnostics, IP scan, and port scan, covering Ping, DNS, Wake-on-LAN, trace route, and more.
• Ports in use: Lists the TCP and UDP ports in use and the process behind each, and can open that folder.
• System info: System, USB, serial ports, printers, sound, fonts, displays, cameras, Bluetooth, disks, and other devices.
• System tools: Read-only lists of processes, environment variables, startup items, installed programs, services, hotfixes, and Store apps, plus file locks and file associations.
• Performance: A borderless window with CPU, GPU, memory, network, and disk, with a pin and an update interval.
• File info: Reads the header type and shows version, image, media, and text-encoding details.
• HTTP request debugger: Requests are saved in groups; parameters are one key=value per line, and a POST can be JSON, XML, or raw text.
• MIDI synthesizer: Play with the mouse or the keyboard, open, change, and save MIDI files, and see a live waveform.
• Sound recorder: Record a microphone, system sound, or both, then play it or save a WAV.
• Camera: Preview a camera, take photos and videos, and scan QR codes and barcodes.
• Scientific calculator: A borderless keypad calculator with trigonometry, logarithms, powers, roots, factorial, and memory.
• Desktop note: A borderless translucent note for text or a to-do list; content and position are remembered.
• RTSP test server: Encodes a bouncing picture to H.264 and serves an RTSP address on the chosen port.
• Auto key: Sends keys or clicks in a row to the foreground window; the default F6 starts and stops it while the window is open.
• Window manager: Lists visible top-level windows and pins one in front or unpins it.

Web tools

• Text
  • Text convert: Encoding, escaping, and common text transforms
  • Chinese to pinyin: Characters to pinyin or initials
  • Markdown reader: Preview Markdown
  • Chinese segmentation: Split by word, sentence, or grapheme
  • In-browser model: Run text tasks on the browser’s local model; Speak can filter voices by language, and the page link opens the Chrome model page
  • Translate: Translate text into another language, with its own engine and local ONNX model pickers
  • Password generator: Generate passwords
  • Regular expressions: Find regex matches
  • Number names: Write a number in several readings
  • Test data: Generate fake data by type
• Math
  • Scientific calculator: Scientific calculation
  • Radix conversion: Convert among base 2, 8, 10, and 16
  • Unit conversion: Length, weight, currency, and more
  • Random number: A number inside a chosen range
  • Function graph: Plot a formula
  • BMI: Body mass index from height and weight
  • Mortgage: Equal payment or equal principal
  • Compound interest: Interest compounded yearly
• Speech
  • Speech: Read text aloud
  • Speech recognition: Turn speech into text
  • 8-bit synthesizer: Square-wave tones and effects, with a live waveform
• Calendar
  • Calendars: Gregorian alongside lunar and other calendars
  • Timestamp: Unix time and local time
  • Date arithmetic: Difference or offset between dates
  • Timer: Stopwatch and countdown
  • World clock: Local time in other places
  • Workdays: Move a date by workdays
• Image
  • Text recognition: Read text in a picture
  • QR / barcode: Make a QR code or barcode
  • Scan QR / barcode: Read a code from a picture
  • Camera: Take a photo or scan a code
  • RTSP player: Play an RTSP video
  • Image Base64: Picture and Base64
  • Image to ICO: Turn a picture into an icon
  • Color picker: Pick or sample a color
  • CSS gradient: Preview a CSS gradient
  • Drawing board: Draw on the page
  • HTML preview: Write HTML and see it
  • Placeholder image: Make a placeholder picture
  • Random colors: A set of random colors
• Lookup
  • Color chart: Named colors
  • This browser: What the browser can read about this device
  • USB IDs: USB vendor and product
  • Unicode chart: Unicode characters
  • ASCII chart: ASCII characters
  • GBK chart: GBK characters
  • Morse code: Morse code
• Talk
  • Chat room: Send text, pictures, and files

Hotkeys

They can be changed in Settings. An empty hotkey string turns that hotkey off.

Hotkey	Action
Ctrl+Alt+O	Show or hide the main window
Ctrl+Alt+Q	Screenshot annotate
Ctrl+Alt+G	Repeat last capture region
Ctrl+Alt+W	Screenshot and recognize
Ctrl+Alt+V	Voice input; press again to finish
Ctrl+Alt+B	Live captions
Ctrl+Alt+T	Show or hide the translate popup
Ctrl+Alt+P	Cycle the screenshot copy mode: image, file, path

Install, language, and updates

The slim package does not include models or the large runtimes; 7za.dll is bundled to extract 7z archives. Screenshot, annotation, recording, Windows OCR, system speech, most tools, file transfer, and the PC side of screencast work right after install; Windows OCR and system speech also need that language in the operating system.

Barcode, ONNX text recognition, offline speech, face, local Chinese–English translation, the dictionary, the input method, PDF, FFmpeg, CUDA, and DirectML are installed or removed under Help → Install features, and using a component that is not installed yet shows a prompt.

Settings → General can turn a group of features off. The UI language is switched on the menu bar or on that same page: Chinese, English, Japanese, Korean, German, Spanish, French, Portuguese, and Russian. Help holds Check for updates, Diagnostics, and Memory.

Run

Windows 10/11 (x64). It needs .NET Framework 4.8.

The local HTTP port defaults to 1224; after Allow LAN access is checked, the same LAN can connect. File transfer, the web tools, and screencast share this port, and POST /api/ocr matches Umi-OCR text recognition.

License

Application source in this repository is under the MIT License. The full text is LICENSE. Copyright (c) 2026 ScreenKit Contributors. Under the MIT terms you may use, copy, modify, merge, publish, distribute, sublicense, and sell it. The software is provided “as is”, with no warranty, express or implied.

The items below do not automatically use this repository’s MIT license. Each license is in THIRD_PARTY_NOTICES.md.

• ONNX Runtime, PDFsharp, PDFtoImage, NAudio, SharpCompress, Vortice, and Bogus are MIT. PDFtoImage also brings the Pdfium and Skia native libraries.
• OpenCV (OpenCvSharp) and ZXing are Apache-2.0.
• 7za.dll (7-Zip) is LGPL-2.1 or later. The license text is License.7za.txt next to the program.
• FFmpeg: the binding code is LGPL-3.0. The ffmpeg64 build itself is GPL or LGPL, depending on that build. Redistribution has to follow those terms.
• DirectML and CUDA / cuDNN follow Microsoft’s and NVIDIA’s redistribution terms.
• ONNX models such as text recognition are usually Apache-2.0, or the terms of that model pack’s author.
• NPinyin turns Chinese into pinyin. The package ships with no license file.
• SQLite (e_sqlite3.dll) opens the input-method database. It is in the public domain.

dict3.bin keeps Chinese, Japanese, Korean, and English in one file. ScreenKit does not build this file. Install features downloads it. Inside it:

• Chinese. Characters, words, and idioms come from mapull/chinese-dictionary (a pinyin dictionary). The license is MIT. The upstream notes say the origin of some material is unclear.
• Japanese. Word senses come from JMdict/EDICT (EDRDG). The license is CC BY-SA 4.0. Redistribution has to keep the EDRDG attribution and follow that project’s update terms.
• Korean. A derivative of the National Institute of Korean Language dictionary. The license is CC BY-SA 2.0 KR.

Step-by-step use, the API, and the build are in the full guide. The phone app is in android/README.md.