Generative ai animation visualizations of the song titles of the WMF cd 'Train to Nowhere'.
For some of the first experiments i was using the term 'hip' in the style prompting with the cultural reference for the word as the intention, but the CLIP embedding was obviously focused on the body part instead. Which again points out that these particular multi-model ai models don't really have a good sense of context and meaning in a sentence, but instead just grab bag associate the individual words in the prompt to image statistics.
No comments:
Post a Comment