Base Tunes

Base tunes are already existing models not fine-tuned to any particular dataset of images. You can use these base tunes directly for prompting or you can create a custom tune by training a dataset on top of them. The following models are available as of Photocobra 1.0.0

i) Realistic Vision v5.1
ii) Realistic Vision V5.0 v5.0 (VAE)
iii) Deliberate
iv) Disney Pixar Cartoon type B v1.0
v) CyberRealistic v4.0

Custom Tunes

As an alternative to base models, you can create a custom tune by training a dataset on top of them. This allows you to fine-tune the model to your specific needs. This new finetuned model can then be used for generating images that are more closely aligned with your dataset (i.e your face).

As a recommendation, we suggest using a highly varied dataset, if you insert the same image multiple times, the model will overfit to that image and will not be able to generate images that are similar to other images in the dataset. This also means that if you are wearing the same shirt in every image the model will overfit to that shirt and will not be able to generate images of you in other outfits. As such, it is important to choose a dataset of images where you are wearing different outfits, have different hairstyles, and are in different lighting conditions.

It is also important for you to be isolated from anyone else in the image, only you should be visible in the image. It is not recommended to use images where you are wearing glasses, hats, or other accessories that cover your face, or images with funny faces.

Face IDs

FaceID is a model adapter designed to generate images that maintain a person's identity without the need for fine-tuning. You can use as few as a single input image. This adapter is specifically trained on human faces and is not suitable for pets or other subjects.

Select model type: FaceID to generate a face ID tune. This can be used in reforge to generate images having this particular face ID. You can also use face IDs to train clothes or shirts or outfits. This can be used later in the reforge to generate images of you wearing outfits trained through face IDs as a means of virtual try-on.