This has example of both low and high resolution data, all from scratch and also accompanying videos! No text to image case though as it only focuses on image modalities.

https://github.com/mikonvergence/DiffusionFastForward