Show HN: Local-First AI Video Upscaler with CPU Fallback (github.com)

A new local-first AI video upscaling tool has been announced, designed to enhance video quality up to 4K resolution while maintaining user privacy. Built in Python, this offline solution utilizes Real-ESRGAN for high-quality enhancements but automatically falls back to FSRCNN for systems without a dedicated GPU, ensuring broad accessibility. The system supports NVIDIA CUDA and Apple Silicon (MPS), and is compatible with macOS, Linux, and Windows, allowing users to upscale videos without any uploads or server-side processing. This tool is significant for the AI/ML community as it prioritizes user privacy and offers a versatile, efficient method to upscale videos, especially useful for restoring personal media or remastering older footage. Key technical features include high-quality Lanczos interpolation for 4K outputs, automatic aspect ratio correction, and original audio extraction, all while being easy to set up with minimal configuration. The project stands out as an empowering resource for individuals seeking to enhance video content responsibly and offline.