Ancestral Seeds are AI generative images and videos trained on the Salaf Dataset. Each white figure or “seed” reveals subtle traces of an orientalist presence. The white figures serve as haunting reminders of old biases, and cultural gaps embedded in our digital landscape. In Ancestral Seeds, the viewer encounters an empty figure — Middle Eastern silhouette echoing the subjects in colonial photography. Yet the expected stereotypes and visual details are missing. By omitting stereotypical features, the work opens a discursive space for viewers to engage with absence as a site of resistance and reinterpretation. Additionally, the generative section illustrates how orientalist caricatures within the dataset — even when stripped down to their most minimal, silhouette forms — continue to persist and reappear in the outputs. This repetition exposes the way AI systems inherit and propagate the visual and ideological residues of the past, reinforcing historical biases through pattern replication. By doing so, the work underscores how algorithmic reproduction is not neutral; rather, it extends and reanimates the colonial.