https://github.com/tripplyons/clip-upscaler-and-enhancer
Using OpenAI's CLIP to upscale and enhance images
https://github.com/tripplyons/clip-upscaler-and-enhancer
Last synced: 11 months ago
JSON representation
Using OpenAI's CLIP to upscale and enhance images
- Host: GitHub
- URL: https://github.com/tripplyons/clip-upscaler-and-enhancer
- Owner: tripplyons
- License: mit
- Created: 2021-12-17T18:28:22.000Z (over 4 years ago)
- Default Branch: main
- Last Pushed: 2021-12-17T19:19:30.000Z (over 4 years ago)
- Last Synced: 2025-04-13T04:51:39.294Z (about 1 year ago)
- Language: Jupyter Notebook
- Size: 3.22 MB
- Stars: 6
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# CLIP Upscaler and Enhancer
Using OpenAI's CLIP to upscale and enhance images
[](https://github.com/tripplyons/clip-upscaler-and-enhancer)
[](https://colab.research.google.com/github/tripplyons/clip-upscaler-and-enhancer/blob/main/clip-upscaler-and-enhancer.ipynb)
Based on [nshepperd's JAX CLIP Guided Diffusion v2.4](https://colab.research.google.com/drive/10YWuTxtBI7PS0xBJCLAUjhR5cB0UUXe-)
## Sample Results
Viewport
Original Image (512x512)
Using This Repo (scaled to 1024x1024 in 45 minutes)
Using Waifu2x (scaled to 1024x1024 in a few seconds)
Full Image



Zoomed In



## How it works
1. Start with any image.
2. Find CLIP's embeddings of the image.
3. Resize the image to the desired resolution and add some noise.
4. **Optimize the resized image for the original image's embeddings** (using nshepperd's JAX CLIP Guided Diffusion v2.4).