Short answer: A memory bot reads values straight out of the game's working memory: exact, fast and technically provable. A pixel bot evaluates only the rendered image and simulates input: slightly less precise, but entirely outside the game process. That difference decides whether a trace exists at all for an anti-cheat system to find.
In this article
How a memory bot works
Every running program keeps its state in working memory: character position, health points, the list of visible objects, inventory contents. These values sit at specific memory addresses.
A memory bot gains access to that memory region and reads the values directly. On Windows this typically works as follows:
- A handle to the game process is requested (
OpenProcess). - Values at known addresses are read via
ReadProcessMemory. - Often a custom DLL is additionally loaded into the process to call or redirect functions directly (hooking).
The advantage is obvious: the bot receives exact numbers. It knows precisely where an enemy stands and how much health it has — no guessing, no misdetection.
The drawback is equally clear. Every one of those steps is an operation the operating system knows about and a protection system can observe. A foreign process requesting a handle to the game and reading its memory is not a grey area. It is an unambiguous event. That is exactly what anti-cheat systems look for.
There is also a practical problem: memory addresses shift with every client update. Memory bots need to be re-adjusted after almost every patch.
How a pixel bot works
A pixel bot skips all of that. It looks only at what is visible on screen anyway and works with classic image processing — in BeaTzVision's case via OpenCV.
Template matching
A previously captured image section, for example an enemy's name bar, is searched for in the current screen image. OpenCV returns a match score between 0 and 1. Above a defined threshold, the target counts as found, including its image coordinates.
Colour analysis in HSV space
For elements more reliably identified by colour than by shape (health bars, status indicators, highlighted names), the image is converted to HSV colour space. The advantage over RGB: hue, saturation and brightness are separated, which keeps detection robust against brightness variation.
Input simulation
The located image coordinates become an action: mouse movement, click, key press. Those events go to the window over the ordinary path any other program uses.
Direct comparison
| Criterion | Memory bot | Pixel bot |
|---|---|---|
| Data source | Game working memory | Rendered screen image |
| Access to game process | Yes, mandatory | No |
| Data accuracy | Exact | Very good, but image-dependent |
| Reaction time | Marginally faster | A few milliseconds slower |
| Provable traces | Yes: handles, hooks, modules | None inside the game process |
| Behaviour on client updates | Usually breaks, addresses change | Keeps working while visuals stay the same |
| Depends on resolution/graphics | No | Yes: templates must match |
| Works across servers | Only with adjusted addresses | Yes, server-independent |
What this means in practice
The row that matters is "provable traces". Every other difference is a trade-off between convenience and effort. This one is categorical: either access to the foreign process exists, or it does not.
The price is real. A pixel bot needs clean setup: templates must match your resolution and graphics settings. Change the resolution and you re-capture. A memory bot does not have that problem. But it stops working after the next client update.
The rule of thumb
Memory bots optimise for precision and pay with detectability. Pixel bots optimise for staying unremarkable and pay with a little setup effort. For Metin2, where sessions run for hours and a single ban costs everything, the second trade is clearly the better one.
How this translates into actual ban risk is covered in detail in Metin2 Bot Without Ban. If you want to get started right away, the setup guide walks you through the first steps.
Frequently asked questions
What is a pixel bot?
A pixel bot is a bot that evaluates only the screen image. Instead of reading values from the game process, it recognises colours, shapes and templates through computer vision and responds with simulated mouse and keyboard input.
What is memory injection in a bot?
With memory injection the bot accesses the game's working memory directly to read values such as positions, health points or object lists. This delivers exact data but requires access to a foreign process, which is loggable.
Is a pixel bot slower than a memory bot?
A pixel bot has marginally higher reaction time because it has to analyse an image rather than read a value. In practice the difference is a few milliseconds and irrelevant for farming.
Ready to try it yourself?
Join the BeaTzTools Discord and get access to BeaTzVision and BeaTzSignal, setup guides and support.
Try it free →