Type | |
Stats | 31 |
Reviews | (1) |
Published | Sep 5, 2024 |
Base Model | |
Training | Steps: 50,000 Epochs: 5,000 |
Hash | AutoV2 D85461DCE3 |
Foreword
The Hypernetwork model transforms a cat into a pseudo-mechanical cat. The cat can become photorealistic, even if the mechanics or the mechanical parts of the cat are surreal.
The grade of photorealism depends on the settings used to create the image. The cat is sitting most of the time, which is based on the fact that I trained the model with sitting cat photos. The background is mostly bokeh or misty also because of the training data.
How to use
For the moment the recommended weighting while using the model is 0.6 to 0.8. Good results are achieved with different models. The quality may vary depending on the model.
Good starting point for an investigations is the Prompt: mechanical cat, <hypernet:pseudo_mechanical_cat:0.7>. Refine the Prompt as required.
Be careful using Hires.fix and a Denoising Strength. This might result in unrealistic an surreal images. But sometimes one can improve an image using Denoising Strength.
Training
The model was created with the help of 76 of my own pictures of sitting cats in front of a blurred background. The resolution used was 1024 x 1024 pixel. I trained the model using AbsoluteReality. Nevertheless, as shown, it works also with a bunch of other models.
Finally
Have fun! Be inspired!