| license: creativeml-openrail-m | |
| tags: | |
| - stable-diffusion | |
| - text-to-image | |
| datasets: | |
| - ProGamerGov/StableDiffusion-v1-5-Regularization-Images | |
| **Object-Taped-To-Wall-Diffusion** | |
| This fine-tuned Stable Diffusion v1.5 model was trained for 2000 iterations with a batch size of 4, on a selection of photos of things taped to a wall. Training was performed using [ShivamShrirao/diffusers](https://github.com/ShivamShrirao/diffusers) with full precision, prior-preservation loss, the train-text-encoder feature, and the new [1.5 MSE VAE from Stability AI](https://huggingface.co/stabilityai/sd-vae-ft-mse). A total of 2100 regularization / class images were used from [here](https://huggingface.co/datasets/ProGamerGov/StableDiffusion-v1-5-Regularization-Images). Regularization images were generated using the prompt "artwork style" with 50 DPM++ 2S a Karras steps and a CFG of 7, using the MSE VAE. A negative prompt of "text" was also used for this dataset. | |
| Use the tokens **ttw style** in your prompts for the effect. Note that the effect also appears to occur at a much weaker strength on prompts that steer the output towards specific artistic styles. | |
| This model will likely not perform well on taping objects that are not traditionally able to be taped to walls. | |
| <div align="center"> | |
| <img src="https://huggingface.co/ProGamerGov/Object-Taped-To-Wall-Diffusion-V1/resolve/main/v1_size_512x512_t4x8.png"> | |
| </div> | |
| * [Full Image](https://huggingface.co/ProGamerGov/Object-Taped-To-Wall-Diffusion-V1/resolve/main/v1_size_512x512_t4x8.png) | |
| Example images were generated with the v1 2000 iteration model using DPM++ 2S a Karras: | |
| ``` | |
| ttw style, <object> taped to wall | |
| ``` | |
| This model was inspired by the 2019 art piece [*Comedian* by Italian artist Maurizio Cattelan](https://en.wikipedia.org/wiki/Comedian_(artwork\)), where a banana was duct taped to a wall. | |