Starting from the five images files that follow,
I will show you how is possible to use AI tools in Fashion design and E-Commerce to edit,
create and brainstorm ideas for your work.
MaxMara Black Cotton Gina Midi dress.
MaxMara Cotton Orange Midi dress.
Max Mara long dress Bridal Pilard gathered maxi gown.
Using a particular model, it’s possible, given an anthropomorphic figure, to make it wear all kinds of clothes we choose
By editing the source image, it is possible to modify and control the final output, as in this case, where I have reduced the length of the skirts.
Here is what the workflow looks like, I am using ComfyUI nodes to control the model parameters,
which in this case is StableDiffusion XL. The DepthMap is extracted through a node and used to maintain the composition.
Using the DepthMap you can create variations and control how much and how the result will be different.
Here I have used one of the latest creations, you can see how it deviates more and more from the original.
The same thing is possible with illustrated material, just change the Model and LoRa to adapt the style.
Training or FineTuning a model throught a “LoRa” makes possible to create infinite variations of colors, materials, and composition as follow :
Few outputs of the model I trained on the Whitney Bag.
Only the text was needed as imput!
Here I combine all together, using the image we created before as reference is possible to have the female model hold the bag we trained the model on.
In this process I started from a open source model like SBxl, and a dataset of the iconic Whitney Bag, scraping the web for good images ( in this case the dataset was of only 15 images ), larger datasets improve the result.
It is important to emphasise that a model such as this can be run locally on one’s own computer, as this is crucial for a company’s copyright and sensitive files.
Jacopo D.G.
Contact@jacopodegiuseppe.com