FramePack
Open source desktop app that turns images and motion prompts into video on Windows or Linux with an NVIDIA RTX GPU.
FramePack is an open source desktop app for making videos from a still image and a written motion prompt. It runs on Windows and Linux, with generation handled by your own NVIDIA GPU. It suits people who want to make AI video locally and see the clip develop as it renders.
FramePack generates video in successive sections. Earlier frames are compressed into a fixed amount of context, so the work needed for each new section doesn't grow with the length of the clip. That approach lets its 13B model generate long videos on laptop GPUs with as little as 6 GB of VRAM. You see the generated sections as they finish, along with a preview of the next section.
The app requires an NVIDIA RTX 30, 40, or 50 series GPU that supports fp16 and bf16. Its models download from Hugging Face, and there is no official hosted FramePack service. The software is licensed under Apache 2.0.
The interface pairs an uploaded image with a motion prompt and shows the resulting video as it grows. FramePack also supports TeaCache to reduce generation time, though the source examples show that it can visibly change the result. Output can vary across hardware even with the same image and prompt.
