https://github.com/hemantkarya/hqstablediffusioncolab
A High Quality (HD / 2K / 4K) Image Generation Using Stable Diffusion and Real-ESR / SwinIR /GFPGAN
https://github.com/hemantkarya/hqstablediffusioncolab
artwork colab diffusion diffusion-models gfpgan hd-images-using-stable-diffusion high-quality-images latent real-esrgan stable-diffusion swinir
Last synced: 3 months ago
JSON representation
A High Quality (HD / 2K / 4K) Image Generation Using Stable Diffusion and Real-ESR / SwinIR /GFPGAN
- Host: GitHub
- URL: https://github.com/hemantkarya/hqstablediffusioncolab
- Owner: HemantKArya
- License: mit
- Created: 2022-09-16T16:56:09.000Z (over 3 years ago)
- Default Branch: main
- Last Pushed: 2022-12-29T09:24:46.000Z (over 3 years ago)
- Last Synced: 2025-04-13T12:45:06.243Z (about 1 year ago)
- Topics: artwork, colab, diffusion, diffusion-models, gfpgan, hd-images-using-stable-diffusion, high-quality-images, latent, real-esrgan, stable-diffusion, swinir
- Language: Jupyter Notebook
- Homepage: https://www.instagram.com/iamhemantindia
- Size: 22.6 MB
- Stars: 35
- Watchers: 4
- Forks: 5
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# **High Quality Text to Image Generation using Stable Diffusion, GFPGAN,Real-ESR and Swin IR**
[](https://colab.research.google.com/github/HemantKArya/HqStableDiffusionColab/blob/main/HighQuality_Text2Image_Stable_Diffusion_ls.ipynb)
[](https://www.instagram.com/iamhemantindia)
Generate 4K and FULL HD Images and Artworks for Free Using Stable Diffusion.
## No Need to generate token key for genrating images from huggingface...
Link to Coalb Notebook [](https://colab.research.google.com/github/HemantKArya/HqStableDiffusionColab/blob/main/HighQuality_Text2Image_Stable_Diffusion_ls.ipynb)
For Upscale Only goto RealESR Notebook (4K Upscale)[](https://colab.research.google.com/github/HemantKArya/HqStableDiffusionColab/blob/main/RealESR_Upscale.ipynb)
Run All the cell until you Reach your Prompt cell.
***In case if you have any human face in your images then it will restore Distorted figures(like eyes,nose,etc) in images, here is example.*** **In Stable Diffusin her Eyes and Lips are bit distorted.**
To upscale images to 2K or 4k using Real-ESR GAN. Note that after running Reasl-ESRGAN leave SwinIR until unless you are not satisfied with RealESR results.
after running this cell you will get a comparison matrix like this.
**Input Images --> Upscaled Images(Real-ESR)**
After Upscaling you images using Real-ESRGAN rest of the cell are optional to run and not recommended (Cause limited GPU RAM in Colab, After running these cell may be it will show you error like ``cuda out of memory``) to run until you are not satisfied with result of Upscaled images of Real-ESR.
right Now I am going to show you difference b/w both Upscalers.
Using both Optional cell at the last of notebook. (It may full your current colab RAM)
**Input Images ------ Upscaled Images(SwinIR) ----- Upscaled Images(RealESRGAN)**
Visit Logical Spot for Video Help:-
[](https://www.youtube.com/c/LogicalSpot)
# **Stable Diffusion** 🎨
*...using `🧨diffusers`*
Stable Diffusion is a text-to-image latent diffusion model created by the researchers and engineers from [CompVis](https://github.com/CompVis), [Stability AI](https://stability.ai/) and [LAION](https://laion.ai/). It's trained on 512x512 images from a subset of the [LAION-5B](https://laion.ai/blog/laion-5b/) database. This model uses a frozen CLIP ViT-L/14 text encoder to condition the model on text prompts. With its 860M UNet and 123M text encoder, the model is relatively lightweight and runs on a GPU with at least 10GB VRAM.
See the [model card](https://huggingface.co/CompVis/stable-diffusion) for more information.
This Colab notebook shows how to use Stable Diffusion with the 🤗 Hugging Face [🧨 Diffusers library](https://github.com/huggingface/diffusers) .
https://github.com/CompVis/stable-diffusion
orignal-link to colab https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/stable_diffusion.ipynb
# **Real-ESRGAN**
[](https://arxiv.org/abs/2107.10833)
[](https://github.com/xinntao/Real-ESRGAN)
[](https://github.com/xinntao/Real-ESRGAN/releases)
[](https://colab.research.google.com/github/HemantKArya/HqStableDiffusionColab/blob/main/RealESR_Upscale.ipynb)
Real-ESRGAN aims at developing Practical Algorithms for General Image/Video Restoration.We extend the powerful ESRGAN to a practical restoration application (namely, Real-ESRGAN), which is trained with pure synthetic data.
# **SwinIR**
[](https://arxiv.org/abs/2108.10257)
[](https://github.com/JingyunLiang/SwinIR)
[](https://github.com/JingyunLiang/SwinIR/releases)
SwinIR achieves state-of-the-art performance on six tasks: image super-resolution (including classical, lightweight and real-world image super-resolution), image denoising (including grayscale and color image denoising) and JPEG compression artifact reduction. See our [paper](https://arxiv.org/abs/2108.10257) and [project page](https://github.com/JingyunLiang/SwinIR) for detailed results.
### (No colorization; No CUDA extensions required)
[](https://arxiv.org/abs/2101.04061)
[](https://github.com/TencentARC/GFPGAN)
[](https://github.com/TencentARC/GFPGAN/releases)
## **GFPGAN** - Towards Real-World Blind Face Restoration with Generative Facial Prior
GFPGAN is a blind face restoration algorithm towards real-world face images.
It leverages the generative face prior in a pre-trained GAN (*e.g.*, StyleGAN2) to restore realistic faces while precerving fidelity.