NotesGPT is an open-source, AI-powered voice note-taking tool that quickly transcribes users' voice notes into text and automatically generates action items. The tool is based on technologies including the Convex database, the Next.js framework, the Together inference model, and more.
MEXMA, developed by Meta AI, is a novel pre-trained cross-lingual sentence encoder. It improves the quality of sentence representations by combining sentence-level and word-level objectives. During training, MEXMA predicts...
Lingua is a lightweight and standalone codebase from Meta AI designed to facilitate large-scale training of language models. Based on easily modifiable PyTorch components, it allows researchers to experiment with new model architectures, loss functions, and...
VideoAgent is a self-improving video generation system developed by researchers from Stanford University, the University of Waterloo, DeepMind, and other institutions. It generates video plans based on image observation and verbal instructions, converting them into robotic...
Llama Tutor is an AI-powered personalized learning platform powered by Llama 3.1 and Together AI technologies. It is a completely open-source project. Users can input their desired learning topics and education level, and the platform generates a personalized learning profile...
ComfyGen, developed by researchers at NVIDIA and Tel Aviv University, is a text-to-image generation system that automatically creates workflows based on Large Language Models (LLMs) to match user-provided text prompts, improving the quality of image generation...
restorePhotos.io is an open-source AI project focused on restoring old and blurry facial photos. It uses the GFPGAN machine learning model to process uploaded photos via Next.js API routing, and then returns the restored photos.
PDFtoChat is an open-source, innovative AI project that allows users to interact with PDF files through natural language dialogue. The tool leverages the latest AI technologies, including Together AI and Mixtral, to understand user queries...
Self-Taught Evaluators are a novel model evaluation method that improves the evaluation capabilities of large language models (LLMs) through self-training, eliminating the need for manually labeled data. Starting with unlabeled instructions, iteratively...
TurboSeek is an open-source AI search engine powered by Together.ai. It is based on a modern technology stack, including Next.js, Tailwind, the Bing search API, and advanced language models Mixtral 8x7B and Lla...
Napkins.dev is an innovative open-source project that uses AI technology to quickly convert user screenshots or wireframes into working web applications. The project is powered by Meta's Llama 3.1 405B large language model and...
Matryoshka Diffusion Models (MDM) is a novel diffusion model framework developed by Apple researchers to address the computational and optimization challenges of generating high-resolution images and videos. MDM is based on...
mPLUG-DocOwl 1.5 is a multimodal large-scale language model launched by Alibaba Group, focusing on OCR-free (optical character recognition-free) document understanding. The model is based on unified structure learning, enhancing its ability to understand rich textual images such as...
Fluid, a text-to-image autoregressive generative model jointly developed by Google DeepMind and MIT, achieves breakthroughs in visual quality and evaluation performance based on continuous labeling and random generation order. The model expands...
SaRA is a novel pre-trained diffusion model fine-tuning method jointly developed by Shanghai Jiao Tong University and Tencent YouTu Lab. It allows the model to adapt to new tasks by reactivating seemingly ineffective parameters from the pre-training process. SaRA is based on...
Quanta Quest is an open-source AI-era intelligent database product for individual users. It integrates personal data from sources like Gmail, Dropbox, and Notion onto a single platform, using RAG technology to provide AI search functionality, helping users...
BiGR is a novel conditional image generation model that uses compact binary latent code for generation training, enhancing the quality and representational power of generated images. As the first model to unify generation and discrimination tasks within the same framework, BiGR...
MoE++ is a novel Mixture-of-Experts architecture jointly developed by Kunlun Tech 2050 Research Institute and Yuan Li's team at Peking University. It is based on the introduction of zero-computational-cost experts, namely zero experts, replication experts, and constant-cost experts...
Granite 3.0 is a suite of advanced AI models from IBM that demonstrates outstanding performance in numerous academic and enterprise benchmarks. Designed for enterprise applications, the models can perform a variety of tasks, including text generation, data analysis, and more.
Project Turntable is an innovative AI tool launched by Adobe at the 2024 MAX conference. The tool allows users to rotate 2D vector images as if they were 3D objects, instantly generating the hidden parts of the image. It uses AI models to fill in...
OmniAI is a powerful AI document processing platform based on advanced OCR technology, supporting rapid processing and data extraction from various file formats such as PDF, DOCX, and images. The platform allows users to easily create document processing pipelines...
TaskWeaver is a code-first AI agent framework from Microsoft, focused on seamlessly planning and executing data analysis tasks. It interprets user requests based on code snippets and efficiently coordinates the execution of various plugins (in the form of functions)...
Stable Diffusion 3.5 is the latest advanced AI image generation model from Stability AI, including Stable Diffusion 3.5 Large, Stable Diffusion 3.5 Large Turbo, and the upcoming...
Mochi 1 is an open-source video generation model from Genmo, known for its excellent motion quality and adherence to user prompts. Released under the Apache 2.0 license, Mochi 1 is free for both personal and commercial use. The model...