This tool allows you to drag and drop your own assets, such as videos, audio, and images, and then use natural language instructions to generate a new video. It uses the Qwen2.5-Coder-32B-Instruct model to process your assets and instructions, to generate a valid FFMPEG command. This command is then executed on your assets to create the desired video.

What's particularly exciting with this is that it's powered by an open-source model licensed under Apache 2.0 (https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct). Tried to build something similar ~1.5 years ago, but at that time, it seemed only possible with proprietary models.