NeFIC is a research-oriented AI model designed for ultra-low-bitrate image compression using video diffusion priors. Developed by a team of authors and presented at ECCV 2026, it focuses on next-frame decoding techniques to address the challenges of compressing images at very low bitrates.
The repository provides six pretrained NeFIC checkpoints, each corresponding to a different rate point, allowing users to select the level of compression suited to their needs. Each checkpoint directory includes two files: weights for the anchor codec and a LoRA adapter for the generative decoder. This setup enables experimentation with various compression rates, from the highest to the lowest bitrate, as indicated by the lambda values provided.
NeFIC is distributed for research-only use, with licensing reflecting this restriction. The model and its checkpoints are available through Hugging Face, and users can download them locally for further investigation or integration into research workflows. Documentation, code, and links to the associated academic paper are provided to support research and development efforts in the field of image compression using AI-driven video diffusion methods.
NeFIC sits in PulseGate's Other AI category. It focuses on compressing images at ultra-low bitrates using advanced AI diffusion models for research and development. It is built as an open-source project for ai researchers. NeFIC is open source under the Open Source license. It runs on the web, API, and the command line.
Yunuo Chen et al. builds and maintains NeFIC, and the product first shipped in 2026. Development happens publicly on GitHub with 1 commits in the last 90 days. Key capabilities include image compression, video diffusion priors, and multiple bitrate checkpoints.
Latest indexed changes and source events
Other apps tracked under the same category.
