Trained upon Microsoft Fluent-Emoji which is described as "a collection of familiar, friendly, and modern emoji from Microsoft"
The Emoji file names were converted to become the text descriptions. It made the model learn a few special words: "flat", "high contrast" and "color"
Description
This is a v2.1 derived version. You will need to copy both the .ckpt as well as the .yaml file to your models/Stable-Diffusion folder
Stable-Diffusion v2.1 fine-tuned for 10k steps using Huggingface Diffusers train_text_to_image script upon Norod78/microsoft-fluentui-emoji-512-whitebg
This model version is also available in both Diffusers as well as Checkpoint formats on its Model card in Huggingface
The inference widget embedded in the model card can be used for quickly trying it out
FAQ
Comments (2)
This is great, thanks! What have you found to be the best configuration/prompts for generating the flat images?
Thank you for the review. In this model I only generated very very simple stuff using rather simple prompts, so other than having the keyword "flat", I'm afraid I don't have any useful info to add. You can take a look at the dataset of Image<->text I created from FluentUI for the purpose of training this if you wish to see "how the model learned" the concepts https://huggingface.co/datasets/Norod78/microsoft-fluentui-emoji-512-whitebg