AutoShorts is a new open-source project for creators who want to process long gameplay footage locally. It uses GPU-accelerated analysis to automatically find action, funny moments, or highlights. The pipeline then crops, renders, and adds subtitles or AI voiceovers without relying on cloud APIs for the core processing.

The tool supports OpenAI and Google Gemini for scene analysis, with a fallback to heuristic scoring if APIs fail. It uses PyTorch and NVENC for fast rendering, and includes ChatterBox TTS for local, multilingual voice synthesis. This makes it a self-contained system for generating short-form content for platforms like TikTok and Reels.

Installation requires an NVIDIA GPU with CUDA and uses a Makefile for setup. By keeping everything local, AutoShorts avoids cloud costs and latency, giving creators more control over their workflow. It represents a growing trend of bringing powerful AI video tools directly to the user's workstation.