Devotd Github
ReleaseMultimodal Runtime Inference

flashml-org/freevideo v0.4.27 — Execution Method and Decoding Recovery

flashml-org/freevideo v0.4.27 is a release that improves the ComfyUI execution flow and the ability to resume decoding after a CUDA error.

Repositoryflashml-org/freevideo ↗

A local inference engine efficiently runs the MiniMax H3 video model on consumer GPUs.

GitHub description: Make videos on the computer you already own. FreeVideo runs MiniMax H3 in as little as 8 GB of VRAM and 16 GB of RAM, and adapts its acceleration path to your hardware.

stars
★ 2.6K
Forks
303
Language
Python
License
Apache-2.0
Latest release
v0.4.27
Read original source ↗View on GitHub ↗

flashml-org/freevideo v0.4.27 was released for Windows, macOS, and Linux. This version provides a flow in which you either select an existing ComfyUI folder or install a fresh one and then run it, and it changed so that decoding can continue even after a CUDA error occurs. The source of these changes is the v0.4.27 release in the GitHub repository.

FreeVideo is a project for making videos on a computer the user already owns. According to the repository description, MiniMax H3 can run with only 8 GB VRAM and 16 GB RAM, and the acceleration path is adjusted to match the hardware. So what this project presents as its core is not to limit itself to one specific piece of hardware, but to choose an execution path that fits the computer you have. The repository description presents the execution requirements, but this information alone cannot be taken to mean that every computer produces the same result.

The changes confirmed in this release are the installation and execution method and the ability to keep decoding after a CUDA error. The conditions under which the error occurs and the specific scope in which resumption works are not given in the available information. The measurement environment for the VRAM and RAM requirements, the processing speed, and the performance comparison with previous versions are not presented either. So v0.4.27 can be seen as an update that adds a path for continuing work after an error, but there is not enough ground to judge the size of the recovery effect or the scope of its application.