Training a model means defining your own parameters based on data you supply. As such that would require a massive dataset and also 100s of H100 Nvidia GPUs to churn that data into model parameters.

This is a simple video showing someone using the latest RTX 5090 to run an existing model and make some AI content at home.

https://cdn.satellite.earth/9d7212700bbeda25189be6f2eab27604e111613b4fdf8f90df72438b446da301.mkv