localai-org/kimodo.cpp

Animate skeletons with natural language; NVIDIA's Kimodo ported to C++/GGML

What it solves

It provides a high-performance C++ implementation of NVIDIA's Kimodo text-to-motion model, allowing users to generate human and robotic character animations from text prompts without needing a full Python environment.

How it works

Built on GGML, the project implements the Kimodo model to translate UTF-8 text prompts or precomputed LLM2Vec embeddings into local rotations and root translations for various skeletons. It supports execution on both CPU and Vulkan (GPU) and includes a DDIM sampling process to generate the motion data.

Who it’s for

Developers and researchers working with character animation, robotics, or game development who need a lightweight, portable way to generate motion sequences from text.

Highlights

  • Multiple Skeleton Support: Compatible with SMPL-X (22 joints), SOMA (30 joints), and Unitree G1 (34 joints).
  • Hardware Acceleration: Supports CPU and Vulkan for inference.
  • Export Capabilities: Can export animations as skeleton-only GLB files for use in tools like Three.js.
  • Flexible Input: Accepts either direct text prompts or precomputed 4096-float embeddings.

Related

  • Project
    kigner/audio.cpp-webui

    A high-performance C++ audio inference framework based on ggml that enables fast, portable local execution of TTS, ASR, and voice cloning models across multiple hardware backends.

  • Project
    mudler/parakeet.cpp

    A C++17 inference port of NVIDIA's NeMo Parakeet speech-recognition models that provides fast, dependency-light ASR on CPU and GPU without requiring Python.

  • Project
    0xShug0/audio.cpp

    A high-performance C++ audio inference framework built on ggml that provides a portable, native runtime for running diverse local audio models including TTS, ASR, and voice conversion.

  • Project
    foldl/chatllm.cpp

    A C++ LLM inference engine based on ggml that enables memory-efficient local execution of multimodal models with RAG and quantization support.

  • Project
    NVIDIA/TensorRT-Edge-LLM

    A high-performance C++ inference runtime for deploying LLMs and multimodal models on NVIDIA Jetson, DRIVE, and DGX Spark edge platforms.