A collection of custom nodes for ComfyUI that adds prompt translation, AI-driven content generation, and image manipulation tools.

An AI-powered tool that uses OpenAI-Whisper and Edge TTS to automatically generate captioned TikTok videos with natural-sounding voiceovers.

Hallo-Live is a real-time text-driven framework that generates synchronized audio and video for digital avatars using a causal dual-stream Diffusion Transformer.

LitData is a Lightning AI library that speeds up PyTorch training by providing fast, cloud‑native data pipelines. It offers two main modes: (1) `StreamingRawDataset` streams raw files directly from S3/GCS/Azure/HF Hub with async, batched downloads; (2) `ld.optimize` converts a dataset into LitData’s own chunked binary format, enabling up‑to‑20× faster streaming, shuffling, and resumable epochs. The library also includes a parallel `map` primitive for preprocessing (e.g., image resizing, embedding creation) and full support for many modalities (images, audio, video, tensors, graphs, etc.). Integration points include PyTorch Lightning, Hugging Face datasets, and Lightning Cloud, making it a practical tool for large‑scale ML projects where I/O is the bottleneck.

Emotion-LLaMA is a multimodal large language model designed for emotion recognition and reasoning by integrating audio, visual, and textual inputs through specialized encoders and instruction tuning.

NagaAgent is a cross‑platform desktop AI assistant (Windows/macOS/Linux) that offers chat with any OpenAI‑compatible LLM, custom tool‑calling via JSON blocks, graph‑RAG memory stored in Neo4j, a Live2D animated avatar, voice synthesis/recognition with lip‑sync, a plug‑in MCP framework for weather, web search, game guides, etc., and community features like a forum and skill marketplace. It ships as an Electron/Vue front‑end plus Python FastAPI back‑end, is installable via `uv sync` (or pip), and is dual‑licensed (AGPL‑3.0 + proprietary).

A diffusion model for generating novel protein backbone structures, enabling the creation of new protein folds for computational protein design.

Clair3 is an open‑source deep‑learning variant caller for long‑read sequencing. It first scans quickly with a pileup‑based neural net, then refines uncertain sites with a full‑alignment model, delivering high accuracy while keeping runtime reasonable. Available as CPU, GPU, and Apple‑Silicon builds, it can be installed via Docker, Singularity, Bioconda, or a Conda environment, and ships pre‑trained PyTorch models for ONT, PacBio HiFi, and Illumina data.

A generative AI movie studio integrated into the Blender Video Editor that allows users to prototype full film productions using AI-generated video, audio, and text.

A local, open-source tool for real-time audio transcription, translation, and OCR, designed for streaming overlays and in-game use.

A ComfyUI plugin for text-to-3D human motion generation based on HY-Motion 1.0, featuring prompt optimization and exports to GLB and FBX.

PointLLM is a multi-modal large language model that enables LLMs to understand colored 3D point clouds for tasks like object classification and captioning.

A vision-language RAG pipeline that embeds documents as images to prevent information loss from text parsing and uses evidence-guided reasoning for multi-image QA.

Diffuman4D is a spatio-temporal diffusion model for high-fidelity 4D consistent human view synthesis, enabling free-viewpoint rendering from sparse-view videos.

NeuMan is a neural radiance field implementation that reconstructs an animatable human and their background scene from a single video, enabling novel view and pose synthesis.

A 3D virtual assistant SDK for Unity that turns 3D VRM models into voice-enabled chatbots with synchronized speech, lip-sync, and autonomous animations.

LiTo is a 3D latent representation system that jointly models object geometry and view-dependent appearance, enabling high-quality image-to-3D generation with realistic lighting effects.

A dependency-free C/C++ implementation of CLIP for efficient inference on memory-constrained devices and serverless deployments.

An open-source Android sample app that demonstrates how to build AI-driven experiences using the Gemini API, Imagen, and Jetpack Compose to rebuild the Android bot maker.

An open-source implementation of a multimodal transformer model that natively processes text, images, and audio within a single architecture.

A ComfyUI extension that automates 3D character rigging and skinning using UniRig and Make it Animatable (MIA) to create animatable meshes.

Tock is an open‑source conversational AI platform that provides an NLP stack, a visual studio for designing dialogue flows, a Kotlin‑based DSL (with Python, Node.js and REST bindings), ready‑made connectors to chat/voice channels, and Docker‑based deployment.

Angel is a Java/Scala‑based distributed machine‑learning and graph‑computing platform that uses a parameter‑server architecture. It supports many classic ML algorithms (LR, SVM, GBDT, LDA*, etc.) and graph algorithms (PageRank, GCN, GraphSAGE, etc.), can run on YARN or locally, and offers a Spark integration (Spark on Angel) for easy job submission.

A PyTorch library for training state-of-the-art multimodal models at scale, providing modular building blocks and pretrained weights for models like CLIP and BLIP-2.

Gemma‑tuner‑multimodal is a macOS‑only Python package that fine‑tunes Google’s Gemma models (3n and 4 series) with LoRA on text, image, or audio data. It runs natively on Apple Silicon (MPS), streams large datasets from GCS/BigQuery, provides a wizard‑style CLI and a live web visualiser, and exports merged checkpoints for Core ML or GGUF inference.

A set of Flutter plugins that enable mobile apps to use Google's standalone ML Kit for on-device vision, natural language, and generative AI features.

A cross-platform AI assistant app for Android, iOS, and macOS that enables real-time chat, multimodal analysis, and the instant creation of web applications using various LLM providers.

A framework and benchmark for developing Multimodal Large Language Models capable of deep research through iterative visual and textual search across images and videos.

Custom ComfyUI nodes that implement Microsoft's TRELLIS.2 model to generate high-quality 3D meshes with PBR materials from a single image.

CC‑Panes is an open‑source desktop app (Windows/macOS/Linux) that lets you run, organise and coordinate multiple AI‑coding CLI agents in parallel. It provides workspace‑level dashboards, reusable launch profiles, an MCP‑based orchestration layer, integrated terminal splits, Git/history tools, planning surfaces, and web/Android access—all built with Tauri, xterm.js and Rust.

DP‑GEN is a Python platform that automates the creation of deep‑learning interatomic potentials (Deep Potential models). It runs a concurrent learning loop—sampling structures, selecting the most informative ones for DFT calculations, and retraining the model—while handling job scheduling on HPC clusters and interfacing with many MD and quantum‑chemistry codes.

An API that integrates ComfyUI nodes into Nuke, allowing VFX artists to use generative AI workflows directly within their professional compositing software.

A Windows desktop assistant that uses vision models to analyze screen content in real-time and generate scrolling AI comments (danmu) with optional voice synthesis.

pytorch‑ood is a genuine Python library for out‑of‑distribution detection built on PyTorch. It bundles >30 detectors, loss functions, pretrained models, datasets and utilities, integrates with pytorch‑lightning, and offers a simple API plus benchmark helpers. Install via `pip install pytorch-ood` and start using detectors like `EnergyBased` with just a few lines of code.

A local-first AI filmmaking workspace that integrates coding agents with a visual canvas and a unified API for generating images, video, and voice.

A spatiotemporal image and video engine that generates AI visuals of any location on Earth at any point in time from 252 million years ago to 3050 AD.

Tutel is Microsoft’s open‑source, high‑performance MoE library for large language models. It provides dynamic parallelism, low‑precision inference (NVFP4, MXFP4, FP8, BF16), support for up to 1 M‑token contexts, vision extensions, and Docker images that serve models like DeepSeek‑V3.2, Kimi‑K3, GLM‑5.x, Qwen‑3, and GPT‑OSS on both NVIDIA and AMD GPUs. Install via pip from the Git repo or build from source, then run the supplied Docker containers or use the Python API for custom MoE routing.

An open-source ecosystem of Chinese pre-trained foundation models and a supporting framework for natural language understanding, generation, and multimodal tasks.

A browser-based studio that combines multitrack video editing and 3D scene direction with AI integration for automated scene composition and editing.

CodexLoom is a self‑hosted platform that adds durable identities, profiles, and organizational tooling to Codex chat agents, enabling long‑running “Domain Agents” that can be accessed from multiple devices, communicate via bounded messages, and be exposed to external chat platforms (Feishu, Slack, Parall) through Interface Agents. Targeted at advanced solo users, it provides UI, CLI, and governance features, and is released under the Elastic License 2.0.

Skywork-R1V is an open-source multimodal reasoning model that uses reinforcement learning and visual chain-of-thought to achieve state-of-the-art performance in complex visual logic, math, and physics tasks.

Torch-RecHub is a PyTorch framework for building recommender systems with 30+ models, modular design, ONNX export, and support for multiple hardware backends.

AnyRecon is a 3D reconstruction framework that uses a video diffusion model to generate high-quality, arbitrary-view reconstructions of scenes from a set of capture views.

A unified multimodal foundation model that integrates image understanding, text-to-image generation, and instruction-guided image editing with a focus on spatial intelligence.

A private, local-first AI companion and art generation platform that combines a customizable chat partner with a layered creative canvas.

An open-source, self-hosted workspace for small merchants to automate the creation of product marketing assets using AI-powered visual workflows for copy and image generation.

A collection of open-source Three.js experiences demonstrating interactive 3D showrooms, games, and data visualizations built with the Mint asset pipeline.

A real-time AI-powered desktop app that detects Bible verses in live audio feeds and renders them as broadcast-ready overlays via NDI.