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.
625 projects
microsoft
Microsoft's large 3D asset generation model (CVPR'25 Spotlight) that turns text or image prompts into high-quality 3D via a unified Structured LATent (SLAT) representation, decoding to Radiance Fields, 3D Gaussians, or meshes. Up to 2B params, MIT-licensed.
m87-labs
A tiny open-source vision language model from m87-labs that runs anywhere, offering captioning, visual question answering, and object detection. Ships a 2B general model and a 0.5B edge-optimized variant under Apache-2.0.
SYSTRAN
A CTranslate2-based reimplementation of OpenAI's Whisper that runs up to 4x faster at the same accuracy with lower memory, adding 8-bit quantization, batched inference, and word-level timestamps. MIT-licensed and FFmpeg-free.
resemble-ai
Resemble AI's MIT-licensed family of state-of-the-art open TTS models, spanning a 350M low-latency Turbo variant and a 23+ language Multilingual V3. Offers zero-shot voice cloning, exaggeration control, and built-in Perth neural watermarking on every clip.
multimodal-art-projection
An open-source, LLM-based foundation model for full-song music generation — a transparent alternative to Suno. A 7B stage-1 model plus a 1B refinement stage turn lyrics into complete songs with coordinated vocals and instrumentals across five languages, under Apache-2.0.
facebookresearch
Meta AI's foundation model for promptable segmentation in images and video, using a streaming memory module to track objects across frames from a single click, box, or mask prompt. Ships Tiny-to-Large checkpoints under Apache-2.0.
jd-opensource
JoyAI-Image is JD.com's open-source unified multimodal foundation model that combines an 8B Multimodal LLM with a 16B Multimodal Diffusion Transformer to handle image understanding, text-to-image generation, and instruction-guided spatial editing inside a single closed-loop architecture. It introduces the OpenSpatial-3M dataset and SpatialEdit benchmark for controllable image manipulation.
zai-org
GLM-V is Z.ai's open-source vision-language model series — spanning GLM-4.1V-Thinking, GLM-4.5V, and GLM-4.6V — that unifies image and video understanding with native multimodal function calling, a 128K token context window, and scalable reinforcement learning for complex visual reasoning. The 106B flagship and a 9B Flash variant are both fully open-source.
QwenLM
Qwen3-Omni is Alibaba Cloud's natively end-to-end omni-modal LLM that jointly processes text, images, audio, and video and generates real-time streaming speech and text responses. It achieves state-of-the-art results on 22 of 36 audio/video benchmarks and open-source SOTA on 32 of 36.
QwenLM
Alibaba's latest Qwen LLM, focused on stability and real-world utility with stronger agentic, repository-level coding and persistent reasoning across turns.
nvidia
NVIDIA's Speech-Augmented Language Model that tops the Open ASR Leaderboard at 5.63% WER, pairing a FastConformer encoder with a Qwen LLM decoder.
OpenBMB
OpenBMB's tokenizer-free TTS system, now a 2B model trained on 2M+ hours across 30 languages with Voice Design, controllable cloning, and 48kHz audio.