Results for “how do diffusion models work

New: try Ask Hacker Search to explore comments!
1.
Photorealistic Video Generation with Diffusion Models (walt-video-diffusion.github.io)156 points, 8 months ago|commentsThis web page discusses a transformer-based approach for photorealistic video generation via diffusion modeling, including text-to-video and image-to-video examples.
2.
Diffusion models from scratch, from a new theoretical perspective (www.chenyang.co)379 points, 4 months ago|commentsThe webpage provides a detailed tutorial on implementing diffusion models from scratch, covering theory, code, training, and sampling. It also discusses denoiser implementation using a multi-layer perceptron and offers further resources for reading.
3.
Diffusion Forcing: Next-Token Prediction Meets Full-Sequence Diffusion (boyuan.space)102 points, 21 days ago|commentsThe webpage describes a new training paradigm that combines full-sequence diffusion models and next-token models for sequence generative modeling, allowing for a range of additional capabilities in tasks such as video prediction and diffusion planning.
4.
Neural Network Diffusion (arxiv.org)223 points, 5 months ago|commentsThis research paper explains how diffusion models can generate high-performing neural network parameters.
5.
Color-Diffusion: using diffusion models to colorize black and white images (github.com)311 points, 12 months ago|commentsThis page explains how color diffusion models work by specifically using them to colorize black and white images.
6.
MobileDiffusion: Rapid text-to-image generation on-device (blog.research.google)261 points, 6 months ago|commentsThis article discusses the efficient latent diffusion model designed for mobile devices for rapid text-to-image generation, including detailed discussions on its background, design, and optimization.
7.
Visual Anagrams: Generating optical illusions with diffusion models (dangeng.github.io)826 points, 8 months ago|commentsThis page discusses generating optical illusions with diffusion models, including examples and conditions required for the illusions to work.
8.
I Made Stable Diffusion XL Smarter by Finetuning It on Bad AI-Generated Images (minimaxir.com)331 points, 11 months ago|commentsThis page provides detailed insights into the AI image generation process, including the technical details of the recent release of Stable Diffusion XL 1.0 (SDXL), how it integrates with the diffusers Python library by Hugging Face, and the use of Dreambooth LoRA for finetuning Stable Diffusion models.
9.
Stable Diffusion in C/C++ (github.com)303 points, 11 months ago|commentsThis page provides detailed information and usage instructions for the stable-diffusion.cpp library, including its features, usage, building and running processes. It includes examples and specific commands for utilizing different functionalities such as downloading weights, building from scratch, quantization, img2img and txt2img examples, and using different models for image processing.
10.
SD4J – Stable Diffusion pipeline in Java using ONNX Runtime (github.com)77 points, 7 months ago|commentsThis page provides detailed information on implementing Stable Diffusion inference using ONNX Runtime in Java, including example images, model support, installation instructions, and implementation details, which is relevant to understanding how diffusion models work.
11.
Extreme video compression with prediction using pre-trainded diffusion models (github.com)144 points, 5 months ago|commentsThe page discusses the use of pre-trained diffusion models for extreme video compression, providing detailed information on usage, benchmarks, and model performance.
12.
How OpenAI's Sora Model Works (www.factorialfunds.com)79 points, 4 months ago|commentsThis technical report dives into the details behind Sora, a diffusion model that builds on top of Diffusion Transformers and Latent Diffusion. It discusses the scaling of video models, implications on GPU inference compute, technical details, implications on synthetic data generation, simulations, and world models.
13.
StreamDiffusion: A pipeline-level solution for real-time interactive generation (github.com)365 points, 7 months ago|commentsThis page details the StreamDiffusion project, including information about real-time interactive image generation. It is relevant to your query about how diffusion models work, as it provides technical details and implementation code for real-time interactive generation techniques.

Terms & Privacy Policy | This site is not affiliated with or sponsored by Hacker News or Y Combinator
Built by @jnnnthnn