Leading artificial intelligence image generation models typically remain closed-weights, whereas video generation models have seen widespread open-weights releases