New CH build tomorrow and side-by-side progress

It's been three weeks since our first weekly build of Catching Heat (CH).  Here's a couple images that highlight the image quality progress that has been made during the meantime.  The first image from version 0.0.1 is blurry without much detail.  The second image, which will be similar to 0.0.3, is clearer with more detail.  The model we are using is a first-generation model and we've already started co-training a second-generation model that experiments have shown to have better quality.  We have our fastest systems training this new model to catch up and surpass the first generation model.  Once that happens, we will switch to the new model for the weekly builds.  Meanwhile, we still have some resources dedicated to refining the first-generation model.

In the background, we are working on a model that will hopefully allow us more control over the details in the images.  This is a new network we have engineered ourselves.  With this new network, we hope to be able to control hair color, finger nail color, clothing and other aspects of the characters.  This new network is very large, however and takes a lot longer to train up to quality so we likely won't see this model being used until quality is equal or better than the current models.