Summary

MTools is a powerful, cross-platform desktop application integrating audio/video processing, image editing, text manipulation, and development tools, enhanced with AI capabilities. It aims to simplify complex workflows and boost user productivity. The application supports GPU acceleration across various platforms, making it a versatile tool for engineers and power users seeking a unified solution for common digital media and development tasks.

How It Works

Built using the Flet framework for a unified GUI, MTools combines diverse functionalities powered by integrated libraries and AI models. Its core advantage lies in performance optimization through GPU acceleration. Users can select from pre-compiled releases or build from source, with specific versions tailored for different GPU acceleration methods, including DirectML (Windows), Core ML (macOS M-series), and CUDA (NVIDIA GPUs), offering flexibility based on hardware and setup preferences.

Quick Start & Requirements

• Recommended: Download pre-compiled executables from the Releases page. No Python installation is needed.
• Source Installation: Requires Windows 10/11, macOS, or Linux with Python 3.11+ and the uv package manager. Clone the repo, run uv sync, then uv run flet run.
• GPU Acceleration: Specific builds are available for Windows, macOS (M-series), and Linux. Source builds require onnxruntime-gpu and potentially manual CUDA 12.x + cuDNN 9.x setup for NVIDIA GPUs.

Highlighted Details

• Cross-platform AI feature acceleration via DirectML, Core ML, and CUDA.
• Multiple build options cater to different GPU hardware and installation preferences.
• Integrates with external AI services (e.g., Atlas Cloud) and models (e.g., ModelScope).
• Comprehensive toolset: image, audio/video, text, and development utilities.

Maintenance & Community

• Maintained by author HG-ha.
• Community support available via QQ group 1029212047.
• Supported by Atlas Cloud and 林枫云.

Licensing & Compatibility

• MIT License.
• Permissive for commercial use and integration.

Limitations & Caveats

• macOS and Linux support are designated as "experimental."
• DirectML builds lack manual VRAM control, unlike CUDA versions.
• Achieving optimal GPU acceleration may necessitate specific build choices or manual configuration of CUDA/cuDNN environments.