# PRIS-CV/DemoFusion

Let us democratise high-resolution generation! (CVPR 2024)

Repository: https://github.com/PRIS-CV/DemoFusion
Canonical: https://ross.abutalabs.com/products/demofusion
Homepage: https://ruoyidu.github.io/demofusion/demofusion.html
Language: Jupyter Notebook
License: MIT
License Family: permissive
Topics: aigc, high-resolution, low-resource, stable-diffusion, genai
Last push: 2025-10-10T01:33:13+00:00

## Health v2 (maintenance only)
Score: 48/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 46, release rhythm 35, longevity 74
- inputs: {"age_days": 1039, "days_push": 328, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2041, forks 216 (observed 2026-08-28T04:06:08.545613+00:00)

## What it is
DemoFusion is a CVPR 2024 framework that extends open-source latent diffusion models like SDXL to generate high-resolution images without additional training or paid services. It uses Progressive Upscaling, Skip Residual, and Dilated Sampling to push resolution beyond the model's native limits on consumer hardware.

## Use cases
- generate 4k images with stable diffusion
- upscale sdxl output beyond 1024x1024
- high-resolution image generation on low-resource gpu
- democratise high-resolution genai without paywalls
- image-to-image generation at high resolution
- use sdxl with controlnet at high resolution

## When to choose
- you need images far above SDXL's native 1024x1024 resolution
- you want to run high-res generation on consumer hardware without paid APIs
- you want progressive previews for rapid prompt iteration
- you want a ComfyUI or HuggingFace-compatible high-res pipeline

## When to avoid
- you need fast single-pass generation - DemoFusion requires many progressive passes
- you need to train or fine-tune diffusion models rather than infer with them
- you need non-image modalities like video or audio
- VRAM is extremely limited even for the low-resolution base pass

## Facets
- artifact type: library
- maturity: stable
- function: stable-diffusion, image-processing, machine-learning, llm-inference
- domain: artificial-intelligence, image-processing, deep-learning, computer-vision
- platform: python, cross-platform
- tags: high-resolution-generation, sdxl, diffusion-models, progressive-upscaling, cvpr-2024, jupyter-notebooks, aigc, gpu

## Member repositories
- PRIS-CV/DemoFusion (main) score 48

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:08.545613+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:58:19.338741+00:00, confidence not recorded.
  - readme: https://github.com/PRIS-CV/DemoFusion (fetched 2026-08-28T04:06:08.545613+00:00, sha 1ebc9d9d41e7)
  - homepage: https://ruoyidu.github.io/demofusion/demofusion.html (fetched 2026-08-29T10:38:30.819804+00:00, sha 0758d6c04511)
- Data as of 2026-08-30T08:39:29.467469+00:00.