Open Source
Explore the latest AI open-source projects from GitHub and HuggingFace.
Explore the latest AI open-source projects from GitHub and HuggingFace.
Explore the latest AI open-source projects from GitHub and HuggingFace. Discover projects across categories like LLM, Vision, Audio, and more.
308 projects
harry0703
AI-powered short video generator that automates scripting, footage sourcing, subtitles, and composition — supporting 10+ LLM providers and batch production.
Fixie AI
Fast multimodal LLM for real-time voice AI that processes audio directly without ASR, built on Llama 3.3 with streaming capabilities.
Crosstalk Solutions
Self-contained offline AI knowledge server with local LLM chat, Wikipedia, Khan Academy courses, maps, and security tools — all running without internet.
HKUDS
Simple and fast RAG framework combining vector search with automatic knowledge graph extraction for dual-level hybrid retrieval across 10+ storage backends.
skypilot-org
Unified framework for running AI workloads across Kubernetes, Slurm, and 20+ cloud providers with automatic cost optimization and spot instance management.
Plastic Labs
Open-source memory infrastructure for building stateful AI agents with continual learning, evolving user models, and multi-provider LLM support.
vllm-project
Extends vLLM to support omni-modality model serving across text, image, video, and audio with high-performance inference.
simular-ai
Open-source GUI agent enabling AI to autonomously use computers with 72.6% accuracy on OSWorld, surpassing human-level performance.
facebookresearch
Meta's self-supervised video learning framework achieving state-of-the-art motion understanding and zero-shot robotics transfer.
ace-step
Open-source music generation foundation model combining diffusion and linear transformers, generating 4 minutes of music in 20 seconds.
QwenLM
Alibaba's most powerful open-source vision-language model with 256K context, spatial reasoning, and visual agent capabilities.
OpenBMB
Tokenizer-free TTS system with context-aware prosody generation and true-to-life zero-shot voice cloning, trained on 1.8M hours of audio.