Pyramid Flow: video generator — what it does and what you need to run it

Pyramid Flow is an open source video generator that tackles the text-to-video task. The model uses a pyramidal approach to flow matching and, as stated by the developers, can produce videos on GPUs with modest memory capacity.

What it does

• Generates video from a text description (text-to-video).
• Supports Multi-GPU mode — using multiple GPUs for acceleration.
• Supports CPU offloading to fit within available VRAM.

The authors describe the model as "Pyramidal Flow Matching for Efficient Video Generative Modeling" and note that it was presented at ICLR 2025.

What you need to run it

• Platform: the developers mention the MPS backend; the code is written in Python, recommended version — 3.8.10.
• Memory: as described by the author, less than 8 GB of GPU VRAM is required. With multiple GPUs and CPU offloading, this figure can be even lower.
• Framework: the model is integrated with the diffusers library, and the weights are available on Hugging Face. The largest weight file is 7.8 GB, with a total size of all files at 27.2 GB.
• License: the code is distributed under the MIT license, while the model weights are under a different license (other).
• Running on your own hardware: possible, the source code is open in the repository github.com/jy0205/Pyramid-Flow. The latest code change is dated December 21, 2024.

Who it's for

The tool will be useful for researchers and developers who want to understand how modern video generative models work or adapt them to their own tasks — especially since the code is open and written in Python using diffusers. The ability to run on GPUs with 8 GB of VRAM and Multi-GPU support make the model accessible to those who lack server-grade accelerators with large memory capacities.

It won't suit those looking for a ready-made cloud solution — "upload text, get video" — without diving into environment setup and working with the command line. It's also worth noting that the model weights have a separate license different from MIT, which may matter for commercial use.

Pyramid Flow is one example of how research models are gradually lowering the entry barrier to video generation. More about the project can be found on the website pyramid-flow.github.io, and the model weights can be downloaded from huggingface.co/rain1011/pyramid-flow-sd3.