SmartEraser is an image editing technology developed by the University of Science and Technology of China and Microsoft Research Asia, specifically designed for removing user-specified objects from images. SmartEraser is based on the innovative "masked-region guidance"...
Zerox is an open-source, localized, high-precision OCR tool based on the GPT-4o-mini model, achieving zero-shot recognition without prior training. Zerox supports various file formats including PDF, DOCX, and images, and excels at handling scanned documents and complex...
Video Alchemist is a new video generation model developed by companies like Snap. It features multi-subject, open-set personalization capabilities and can generate videos based on text prompts and reference images without requiring optimization during testing. The model is based on...
PSHuman is an advanced single-image 3D portrait reconstruction technology. Based on a cross-scale multi-view diffusion model, it can generate highly realistic 3D portrait models, including detailed facial expressions and full-body poses, from just one photograph. Its core advantage lies in...
Wear-Any-Way is an innovative virtual try-on framework launched by Alibaba's Taobao Image Search team. Through a sparse correspondence alignment mechanism, it achieves high-fidelity and customizable virtual try-on effects. Users can generate realistic try-on images,...
GameFactory is an innovative framework jointly proposed by the University of Hong Kong and Kuaishou Technology to solve the scene generalization problem in game video generation. The framework is based on a pre-trained video diffusion model, combined with open-domain video data and small-scale high-quality...
Step-2 mini is a lightweight, high-speed large model launched by Step-Star, developed based on the next-generation self-developed Attention architecture MFA. It retains over 80% of the performance of Step-2 with only 3% of the parameters, significantly improving performance...
FramePainter is an AI-based interactive image editing tool that combines a video diffusion model with intuitive sketch controls, allowing users to indicate their editing intentions through simple drawing, clicking, or dragging operations, thus enabling them to manipulate images...
Bailing is an open-source voice dialogue assistant that uses Automatic Speech Recognition (ASR), Voice Activity Detection (VAD), Large Language Model (LLM), and Text-to-Speech (TTS) technologies to achieve natural voice dialogue with users, realizing a near-universal...
LineArt, developed by Jilin University, the Royal Institute of Technology in Sweden, and the Tokyo Institute of Technology, is a high-quality design drawing appearance transfer framework that requires no training. It can transfer complex appearance features to detailed design drawings, assisting designers in...
SynthLight is a portrait relighting technology developed jointly by Yale University and Adobe Research. Based on a diffusion model, it is trained using synthetic data simulated under different lighting conditions to re-render portrait photos with...
X-Dyna is an animation generation framework based on a diffusion model. It animates single human images by driving facial expressions and body movements in videos, generating dynamic effects with realism and environmental awareness. Its core is Dynam...
DeepSeek-R1 is a high-performance AI inference model developed by DeepSeek, a company based in Hangzhou. It is comparable to OpenAI's official O1 release. The DeepSeek-R1 inference model is post-trained using large-scale reinforcement learning techniques, requiring minimal...
k1.5 is the latest multimodal thinking model launched by Moonlit Dark Side Technology, possessing powerful reasoning and multimodal processing capabilities. Under the short-CoT (short-chain thinking) mode, the model exhibits significant advancements in mathematical, code, and visual multimodal capabilities, as well as general-purpose abilities...
OmniThink is an innovative machine writing framework jointly developed by Zhejiang University and Alibaba's Tongyi Lab. By simulating the iterative expansion and reflection process of humans, it breaks through the knowledge boundaries of large language models in machine writing. The framework...
H-Optimus-0 is the world's largest open-source AI model for pathology, released by the French startup Bioptimus. The model has 1.1 billion parameters and was trained on a proprietary dataset containing 500... from 4,000 clinical practices.
Roop-Unleashed is an open-source project based on Roop, focusing on the implementation and optimization of deepfake technology. Users can quickly replace faces in images and videos without complex training processes. ...
The "Shusheng·Puxiang" algorithm is an ultra-high dynamic range imaging algorithm developed by the Shanghai Artificial Intelligence Laboratory in collaboration with the Chinese University of Hong Kong, Zhejiang University, and other institutions. Its core is Puxiang HDR (UltraFusion HDR). The algorithm combines AIGC technology and HDR technology...
DeepSeek R1-Zero is an inference model developed by the DeepSeek team that relies entirely on pure reinforcement learning (RL) training, without using any supervised fine-tuning (SFT) data. It performs exceptionally well on inference tasks, achieving high scores in AIME 2024...
VideoWorld is a research project jointly conducted by Beijing Jiaotong University, the University of Science and Technology of China, and ByteDance. It explores whether deep generative models can learn complex knowledge, including rules, from unlabeled video data alone.
Step-1o Vision is the visual version of Step-1o's latest native end-to-end multimodal generation and understanding integrated model. Focusing on visual tasks, it boasts powerful image recognition, perception, reasoning, and instruction following capabilities...
3DIS-FLUX is a deep learning-based multi-instance generation framework that achieves high-quality image generation through decoupled instance synthesis. Combining the depth-driven scene construction of the 3DIS framework and the diffusion transformer architecture of the FLUX model, it is implemented in two stages...
DITTO-2 is a novel music generation model jointly developed by Adobe and researchers at the University of California. It achieves fast and controllable music generation by optimizing the inference time of a diffusion model. The model is based on optimizing the inference time of a diffusion model...
DiffEditor is an image editing tool based on the diffusion model, proposed by a research team from Peking University Shenzhen Graduate School and Tencent PCG. It introduces image prompts and text hints...