Capture an image
The game window is captured over and over. The bot waits until the image is stable before reading it.
Nothing here is secret. GMI L2M reads an image of the game window, then sends the same key presses and clicks you would make yourself. This page explains each part, including its limits.
The game window is captured over and over. The bot waits until the image is stable before reading it.
The detector recognizes the state of the screen, image matchers look for buttons and panels, and the text reader reads dialogs.
The order of the checks is fixed, and their content comes from your settings: which key, when, at what HP.
Key presses are sent to that game window. Clicks use the Windows cursor, with accounts taking turns.
Every account on one PC shares two things: the detector that reads the images, and the Windows cursor. Both are used in turn, and that is what decides how many accounts are reasonable for your PC.
This is an explanation of how it works, not a guarantee. Automation can violate the game's terms of service, and the publisher can impose sanctions on the account. No way of working removes that risk.
The game changes, clients differ, unexpected screens appear. What matters is how quickly you learn what happened.
Every wait for a screen is logged: what it was waiting for, how long, and the result. Past 20 seconds, a diagnostic bundle is created automatically.
The Recordings page reads that bundle and writes up what it finds: which screen was open, which matcher almost matched, and a suggested action.
The most common cause is a reference image that does not match your client. Capture it again from your own screen, then try again.
Read the guide, check the status of each feature, and decide in your own time.