This paper examines how generative Artificial Intelligence (AI) can contribute to the artistic production of abstract images for video art, with an emphasis on human-machine collaboration through natural language instructions. It starts by examining how artists incorporate AI into their practices, challenging traditional notions of authorship and leveraging the unpredictability of generative systems to benefit the creative process. The study employs the performative research paradigm, which is adaptive and encourages free exploration. This approach is reflected in the practical component, which involves iteration and experimentation – from prompt engineering to parameter optimization – to explore the models’ capabilities. The aim is to create abstract images for a video art essay in collaboration with the Stable Diffusion and Midjourney models, using instructions derived from terms that describe abstractionism and its relationship with moving images. The findings, presented also in the form of a video essay, reveal that, although generative AI plays an active role in establishing a visual foundation, creativity emerges from the human interaction with the systems, primarily through text-based instructions and decision-making regarding the produced images.