Kimodo CPP
A C++ port of Kimodo contributed by developer Richard Pellentropi. It enables the Kimodo animation generation model to run on CPU-only hardware or devices with as little as 2 GB of VRAM, substantially lowering the hardware requirements compared to the original implementation.
Timeline (last 90 days)
Kimodo CPP is a C++ port of Kimodo contributed by developer Richard Pellentropi, enabling it to run on pure CPU or devices with very low VRAM
Kimodo CPP requires as little as 2GB of VRAM to run and is not as sensitive to increases in VRAM as the original version
Kimodo CPP takes approximately 5 minutes to generate a 3-second animation on pure CPU or with 2–4GB of VRAM
Kimodo CPP still requires a minimum of 4GB of RAM, and the model weights themselves occupy approximately 25GB of storage space
Kimodo CPP is hosted on GitHub and designed for Linux by default; Windows installation dependencies include C++ Build Tools, Vulkan SDK, Golang, and Git
Running Kimodo CPP requires downloading the Kimodo GGML weights for Llama 3 as well as the skeleton model used to drive it
The initial release of Kimodo CPP only supports the SMPLX skeleton, and its license is a non-commercial research license that does not permit commercial use
Kimodo CPP currently supports four skeletons; in addition to SMPLX, Soma and G1 skeletons are also available
Kimodo CPP defaults to 20 inference steps, supports both GUI and command-line operation, and can be forced to use CPU computation via a parameter
The latest version of Kimodo CPP natively supports exporting GLB format in the UI, whereas earlier versions could not export GLB
1 more timeline events
All Facts (11)
Kimodo CPP is a C++ port of Kimodo contributed by developer Richard Pellentropi, enabling it to run on pure CPU or devices with very low VRAM
50%UnverifiedKimodo CPP requires as little as 2GB of VRAM to run and is not as sensitive to increases in VRAM as the original version
50%UnverifiedKimodo CPP takes approximately 5 minutes to generate a 3-second animation on pure CPU or with 2–4GB of VRAM
50%UnverifiedKimodo CPP still requires a minimum of 4GB of RAM, and the model weights themselves occupy approximately 25GB of storage space
50%UnverifiedKimodo CPP is hosted on GitHub and designed for Linux by default; Windows installation dependencies include C++ Build Tools, Vulkan SDK, Golang, and Git
50%UnverifiedRunning Kimodo CPP requires downloading the Kimodo GGML weights for Llama 3 as well as the skeleton model used to drive it
50%UnverifiedThe initial release of Kimodo CPP only supports the SMPLX skeleton, and its license is a non-commercial research license that does not permit commercial use
50%UnverifiedKimodo CPP currently supports four skeletons; in addition to SMPLX, Soma and G1 skeletons are also available
50%UnverifiedKimodo CPP defaults to 20 inference steps, supports both GUI and command-line operation, and can be forced to use CPU computation via a parameter
50%UnverifiedThe latest version of Kimodo CPP natively supports exporting GLB format in the UI, whereas earlier versions could not export GLB
50%UnverifiedAnimations generated by Kimodo CPP currently have no finger animation, and the hands are essentially in a relaxed pose
50%