Everything measured on one Mac Studio, the mothership
On a Mac Studio (M5 Max, 36GB),how fast can it
make videos?
- Memory
- 36GB
- Chip
- M5 Max
- CPU
- 18 cores
- GPU
- 32 cores
The mothership: Mac Studio (model MHL64J/A), macOS 27.0. Every number on this site was measured on this one Mac. Whether an AI runs at all comes down mostly to how much memory you have.
AI that makes short videos can also run inside your own Mac. It takes far longer than images, though, and uses a lot of memory. On this page, video AI models are given the same prompts on the mothership Mac Studio, and the time they take and the results are compared.
Results by model
- Wan 2.2 TI2V-5B by Alibaba (Wan): about 14 min for one 3.4-second clip, 45.7GB of memory
To try it on your own Mac (Draw Things)
- Install Draw Things. Search the Mac App Store for “Draw Things” and install it (free). It is the same app as for images, and you don't need Terminal.
- Pick a video model. Choose a video model from the app's model list, for example one with “Wan 2.2” in its name. The first time, the model downloads. The file is large, so this takes a while.
- Start with a short clip. Video takes far longer than an image. At first, lower the number of frames to about 14–25, write a short description of the clip you want in the text box (the prompt) and press the generate button. Descriptions in English are followed more closely.
- Save it. The finished clip stays in the history. Right-click it there to export it as a video file.
These steps follow Draw Things' own guide (Video Generation Basics). That guide suggests 8GB of memory or more for video, and 20GB or more for the large models. The measurements on this page use a different tool (mlx-video), so the time and memory in Draw Things will not match the numbers here. Draw Things can also generate in the cloud; to keep everything on your Mac, don't use that feature and download the model instead.
How it was measured
- Tool: made with mlx-video, measuring the time to finish one clip (including loading the model) and the most memory used meanwhile.
- Made with the same random number (seed) for every prompt.
- Only models under a license that lets their videos be published freely (Apache 2.0) are shown.
- Release date: the day the model appeared on Hugging Face (when its repository was created). It can be a little earlier than the announced date.