Local OCR chat overlay that judges intent and drafts replies via LLM.
Offline OCR screen-reader feeding LLMs to generate contextual chat response options.
This is a Windows desktop overlay that reads chat windows via local OCR, evaluates intent, and generates three candidate replies for you to insert manually. It suits professionals and power users managing high-volume desktop chats who want AI-assisted drafting without auto-sending.
It captures your active chat window using Windows Graphics Capture and runs RapidOCR locally to parse the conversation. A dual-model backend uses one LLM to judge intent and context, and a second LLM to draft three scored reply options that you can paste with a single click.
It suits Windows power users looking for context-aware reply suggestions across chat applications without hooking into app processes. It does not suit non-Windows users, those seeking automated message sending, or anyone unwilling to supply their own API keys.
Captures screen regions and parses text offline using RapidOCR without touching app databases.
Separates intent judgment from response drafting to ground suggestions in actual conversation flow.
Fills the target chat input box with your chosen candidate while keeping sending strictly manual.
Feeds your recent short messages as style samples so drafts mimic your personal writing tone.
Parses speaker names and allows targeting specific participants for tailored responses.
The software is open source and free to download, but you must pay the API usage costs for the external LLM services you configure for judgment and drafting.
874 open remote AI-training roles, updated November 15, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
These reviewed AI tools do the same job instantly — from our scored directory.
No, it only populates the chat input box. Sending is always performed manually by you.
API keys are written directly to your Windows user environment registry variables and never saved to local config files.
It relies strictly on screen capture and local OCR, functioning identically to standard screen-reading software rather than injecting code.
See Zekai first in Google