Step 1: Prepare Images

Step 1: Prepare images

Gather various images – for example for this tutorial we will pick 30 or so images of a lady. The images must be from a good camera and without use of any filters. If you do that, you will get their likeness a lot better.

It’s important that you choose Head Shots, Partial Body Shots, Full Body Shots, some Close-ups, and also some Far Away shots.

You need to ensure that; (a) The Hairstyle is different in images, (b) The Photos are taken from different profiles, (c) The Backgrounds change, and (d) Outfits change.

If you pick too many head shots into your training data, you will end up with poor results. This is because the AI doesn't know what the ‘human body’ is, so it gets confused. You absolutely need to put some body shots in the collection.

Rize and Rename your images. You can use a website called BIRME (Bulk Image Resizing Made Easy) ver. 2.0 Software (https://www.birme.net), which is free to use. BIRME is used to crop the images down to the size you want.

We are going to be using 512 x 512 because that is pretty much a standard resolution. You can go up to 768 by 768, but be aware that higher resolution will use more VRAM and take longer to train the images.

After resizing, rename all the files and save the images collection as a ‘zip’ file (use any file name such as ‘myimages.zip’). The files will be saved in numerical order i.e. 1.png, 2.png, 3.png, 4.png etc. etc. Next, upload your zip file to your Google Drive