CAPTION · OCR · SUBTITLES
gaze
GAZE · @JIANGXI1129
Lend the AI a controlled eye and let it know what you are looking at together.
gaze uses a native Python loop to capture the window explicitly selected by the user, and then writes the screen overview, OCR text, and audio subtitles into a scrolling context that can be read by the AI. It emphasizes narrow-scope capture, privacy blacklists, and recursive protection that prevents the AI from seeing its own interface.
Project address (can be copied to AI):https://github.com/jiangxi1129/gaze
PROJECT INTRO
Project introduction
gaze is a native tool for adding screen awareness to existing AI companions. It can continuously capture a specified window or a controlled foreground window, use a visual model to generate short narrations, use OCR to extract text, and optionally listen to system audio to generate subtitles. Real-time context can be written to simple JSON, JSONL, or any own memory service.
The project adopts a narrow-gate strategy by default: if the target window cannot be found, it will be skipped, and the entire desktop will not be automatically intercepted; windows such as banking, chat, password manager, and email are blacklisted, and at the same time, AI clients and floating windows are independently prevented from being recursively captured. Users can also record completely offline, without sending footage to external services.
- Suitable for:People who want AI to accompany them to watch videos, read screens, and play games while retaining clear visual boundaries.
- Platform:Windows 10/11; pure Python, minimal mode without Node or browser automation.
- Usage:Select Window and Visual Services, and enable OCR, audio subtitles, and memory push as needed.
- License:MIT。
COMMENTS & FEEDBACK
Comments and Feedback
People who have used this project can come back and tell newcomers: which platform it ran on, whether there were any pitfalls during the installation, and what the actual experience was like. The current version is in guest mode, and you do not need to register an account to leave a message.
The message is waiting for review and will be publicly displayed here after approval.
SOURCE & CREDIT
Atlas is responsible for the introduction and navigation, and the use still returns to the original author.
We do not mirror project files, nor do we intercept author traffic. For installation, download, version updates and the latest instructions, please refer to the project page provided by the original author.
Reading public comments...