The filmmaker who
actually directs machines

Every frame begins as language. A sentence becomes a scene, a phrase shifts the light, a single word redirects the camera. Making films with generative video is a dialogue with probability—you propose a world, the model interprets it, and the negotiation between intent and emergence produces something neither party could have authored alone. These films are not demonstrations of technology. They are the output of a creative practice that treats foundation models as instruments—each with its own grain, its own rhythm, its own way of seeing.

The workflow operates across the full model stack. Text-to-video generation begins with Kling v3 Pro and Seedance 2.0 for primary scene creation—their native audio generation and multi-shot editing capabilities allow narrative structure to emerge directly from the model. Veo 3.1 and Sora 2 handle passages requiring higher temporal coherence and precise camera control. Image-to-video pipelines using PixVerse V6, Wan 2.2, and LTX Video 13B extend still compositions into motion—animating concept frames, key art, and photographic references into cinematic sequences. Each film typically moves through 200–400 generations before the final cut materializes.

The work gravitates toward liminal spaces, memory textures, and the uncanny beauty that emerges when models hallucinate with purpose. There is a specific quality to AI-generated movement—a fluidity that sits between photography and painting, between documentary and dream—that becomes the visual signature across these films. Each model introduces its own artifacts and tendencies: Kling’s cinematic depth of field, Seedance’s physics-aware motion, Veo’s compositional precision, Sora’s temporal drift. Rather than suppressing these qualities, the practice amplifies them—using model behavior as a creative material in itself.

Each new model generation collapses the gap between prompt and vision. Multi-shot editing, native audio synthesis, and reference-to-video pipelines have compressed what was once a week of iteration into a single session. The constraint is no longer the tool—it is the clarity of the creative intention behind the prompt. The films on this page represent a practice in constant motion: learning to work with systems that are simultaneously camera, actor, set designer, lighting technician, and cinematographer. The filmmaker’s role has not disappeared. It has shifted—from controlling every frame to shaping the conditions under which frames emerge.