I've changed up some of the parameters and changed the dataset. I'm still tweaking a few things. This model is only half baked. I had to stop because I wanted to tweak more things and see if it could be even better. Still working on a final model.There isn't a specific trigger word, it is trained to use natural language, though you have to keep it sort of simple.learned concepts:<ul><li>nude man</li><li>penis (may also know cock and dick)</li><li>*erect penis - only sometimes (may also know "hard")</li><li>*cum/ ejaculating cum (may also know semen and shooting stream of semen) - hit or miss</li><li>point of view</li><li>*masturbating - only sometimes</li><li>penis through the fly</li><li>*flaccid penis - isn't really needed usually</li><li>point of view photo of a man inserting his tongue into the foreskin of the viewer's erect penis* is not quite as good as my other tongue in foreskin models unfortunately</li></ul>

Files with SHA256: e0b3c8c54a4a013a2dbb8897f322ca6e04f236506cc29ead4af0bb14af33c81f