Following the emergence of AI image generators last year, the introduction of #Sora last week gives us now access to text-to-video generators. Would you guess what our take is on this?
"Aftermarket" fixes applied after training, like injecting diversity terms into prompts, don't fix the underlying model and can even exacerbate harmful fabrications. If the training set is biased—and the Internet is—it's really hard to correct that after the fact.
prompt: 'abstract dog t shirt design. Deconstructed minimalist line drawing. Fractal split effect. High fashion style of Dolce Gabbana and Balenciaga --chaos 12 --ar 1:1 --style raw --stylize 750'