GPTBuster is a real-time LLM and imageGen detection tool designed to help educators and institutions identify AI-generated text or images directly from screen content. Triggered via a hotkey, it performs local, high-certainty analysis using a high-performance Rust backend.
GPT Buster accesses window content independent of the CPU via Direct Memory Access (DMA). This setup allows for real-time AI Detection with no performance impact on the Host PC.
GPTBuster is now fully open source under the Apache 2.0 License, enabling trusted offline deployment.
⚠️ No data ever leaves your system — 100% local processing.
- ⌨️ Trigger detection with a hotkey
- 🧠 Detect AI-generated text and images
- 🛡️ Entirely offline and private (Rust-based backend)
- 🖥️ Works via Direct Memory Access (DMA) screen capture
- 🧩 Offers optional WebSocket API via gptbuster.com
- ⚙️ Open-source with Apache 2.0 License
- ⚡ Requires a high-end GPU (≥32GB VRAM) for real-time performance
GPTBuster uses a 2-computer setup:
- Displays and interacts with content
- Sends raw display memory to Analyzer Host via DMA
Minimum Requirements:
- CPU: Any modern CPU (Intel i5/Ryzen 5 or higher)
- RAM: 16GB
- GPU: Integrated or entry-level discrete
- Supports DMA output (via compatible capture card)
- Performs real-time detection and processing
Minimum Requirements:
- CPU: 12+ cores
- RAM: 64GB+
- GPU: 32GB+ VRAM (e.g., RTX 6000 Ada, A100, H100)
- Storage: NVMe SSD recommended
- DMA Card for memory capture
- Fuser to mirror display to monitor
GPTBuster uses a Direct Memory Access (DMA) card to capture screen memory directly from the Capture Host, bypassing the CPU and OS. This enables low-latency, high-bandwidth transfer to the Analyzer Host.
A fuser device allows simultaneous screen display and memory capture, ensuring the user experience remains seamless.
⚠️ GPTBuster is intended for advanced setups with access to enterprise-grade GPUs. Use only in environments with sufficient hardware.
-
Clone the repo:
git clone https://github.com/gptbuster/gptbuster.git cd gptbuster -
Build the project (requires Rust nightly):
cargo build --release
-
Configure hotkey and capture settings in
config.toml -
Run the analyzer:
./target/release/gptbuster
Don't have the hardware? GPTBuster offers an enterprise-grade detection API with the same engine:
- WebSocket API hosted at: gptbuster.com
- Ideal for schools without access to 32GB+ VRAM GPUs
- Secure, fast, and supports both text and image detection
This project is licensed under the Apache 2.0 License.
See LICENSE for details.
We welcome contributions from universities, researchers, and developers!
- Fork the repo
- Submit pull requests
- Open an issue to discuss major features