xushengfeng/eSearch

What it solves

eSearch is a cross-platform screen search tool designed to bring mobile-style "screen search" (similar to Smartisan Big Bang or Xiaomi Portal) to desktop environments. It combines screenshotting, text recognition, and information retrieval into a single workflow to eliminate the friction of manually typing out text seen on screen.

How it works

The application runs in the system tray and is triggered via a shortcut. It uses a combination of local and cloud-based AI models to process visual information:

• OCR: Uses an offline engine based on PaddleOCR (via eSearch-OCR) for local text extraction, with options for online OCR via Baidu and Youdao.
• Translation: Integrates various translation engines (Google, DeepL, Baidu, ChatGPT) to translate recognized text.
• Visual Search: Connects to Google, Baidu, and Yandex for reverse image searches.
• AI Integration: Supports custom multimodal LLM interfaces to allow users to "chat" with the images they capture.

Who it’s for

Users on Windows, Linux, and macOS who need a high-productivity tool for extracting text, translating foreign languages, or identifying objects within images directly from their screen.

Highlights

• Integrated Workflow: Seamlessly moves from screenshot $\rightarrow$ OCR $\rightarrow$ Translation $\rightarrow$ Search.
• Offline OCR: Built-in offline text recognition supporting multiple languages and vertical text.
• Advanced Screen Capture: Includes scrolling screenshots, a color picker, and image beautification tools (shadows, rounded corners).
• Screen Translation: Can replace text on the screen with translated text in a floating window, useful for games or videos.
• Screen Pinning: Allows users to pin captured images to the screen with adjustable transparency and scaling.
• Screen Recording: Features a "Super Recording" mode that automatically zooms and focuses on the mouse cursor.