FramePack

FramePack is a desktop app for creating AI video from a still image and a written motion prompt. It's for people who want to animate images on their own hardware and inspect the clip while generation is still running.

• Image animation: Describe subject movement or camera motion to turn a still image into a video.
• Progressive previews: See completed video sections as they accumulate, plus a preview of the next section before it finishes.
• Longer clips on smaller GPUs: Generate minute-long videos with a 13B model using as little as 6 GB of VRAM, including on laptop GPUs.
• Faster drafts: TeaCache reduces generation time, but it can visibly change the result, so it suits testing ideas before running the full generation process.

FramePack runs locally on Windows and Linux. It requires an NVIDIA RTX 30, 40 or 50 series GPU with fp16 and bf16 support and at least 6 GB of GPU memory. The software is open source under Apache 2.0.

Generation can take substantially longer on laptop GPUs than on desktop hardware. The same image and prompt can also produce slightly different results across devices.