You can customize models to teach it new concepts. New concept can be an Object, Style or a Human. You need a few images to train a new concept to the model and then you can generate images where you can imagine Object in different places or imagine anything else in a particular Style.
Training a Custom Model
- Click on Train your own button on homepage.

- Give a name to your model that represents what you are trying to train. Since we are training cat and cat is an object, we used Object in our example. Then click on Upload Images and Train button.
- Upload images of Style/Object/Human that you want to use for training.
Upload at least 8 images and on average 15 images.
β1. Object/Human in focus.
β2. Different context and lighting.
β3. Only contains the Object/Human - Try cropping images before uploading.

Below are 11 images of black cat we used to train our model :

- Once images have been uploaded it will take ~20 mins to train your model. You will receive an email when its done. You can then use this custom model to train images.
Generating from a Custom Model
- On home page, click on Personal Model button. You can then choose which Custom Model you would like to use for generation from list of all pre-trained models. A screenshot of this step:
- There is a special keyword "X" Β in capital which will relate to Object/Human/Style you trained.

- Let's say you trained an object of your cat. Β Remember to use "X" to correspond to your object in prompt. Here are generations we got from below prompt:
X on a cruise sitting and enjoying a cocktail with sunglasses on

Training a Human
Now, lets look at how to your images, any human or character. We are going to train "wednesday addams" which is a character from a popular series. Below are input images used to train model by clicking Human asw training subjct.

Below prompt was used to create images from this model in portrait dimensions:

Training a Style
Now, lets look at how to train style like abstract, minimalism, pop. We are going to train illustration style. Below are input images used to train model by clicking Style as training subject.

Below prompt was used to create images from this model :
Summer nights in X style

Please send us a message if you find anything confusing and we will help you out.