Create Cinematic AI Films in 1 Hour: Workflow & Tools

The filmmaking process begins by focusing on a story with real emotional stakes, rather than just visually appealing concepts. The goal is to tell a story that resonates deeply with the audience. For the example film, the core idea was a depressed cop on his birthday, with loved ones attempting to cheer him up, condensed into a three-sentence foundation.

From this foundation, the narrative is mapped into three distinct scenes: a quiet locker room where the partner notices something is off, a casual yet tense patrol car conversation where the partner tries to uplift him, and a surprise moment that shifts the entire dynamic. This structure provides a clear setup, midpoint, and climax, which is sufficient for a short film.

To create a consistent character, the first step involves uploading 20 photos of the subject (the filmmaker) to Hicksfield's character selection, ensuring a variety of angles and lighting. This process trains an AI model to recognize the face, generating a personal Soul ID that can be applied to any scene for consistent appearance.

Soul Cinema is then used to generate initial cinematic images, such as a hero shot of the cop. To maintain visual consistency across different shots and animations, this image is taken into Nano Banana Pro to create a detailed character sheet. This sheet locks in every detail, including face, accessories, and clothing, preventing visual drift. The same process is applied to generate and define supporting characters, like the partner Dave, ensuring their faces are production-ready with natural color grading and skin texture.

For scene one, a locker room environment is generated using Soul Cinema with a simple prompt like 'A wide shot of a locker room with blue lockers.' The blue lockers are chosen for their cold, institutional feel, emphasizing the character's isolation. A prop sheet is also generated using the same prompt to ensure environmental consistency across all shots within the scene.

This prop sheet helps maintain the integrity of the environment, ensuring that elements like lockers and other background details remain consistent, even when generating multiple shots or different camera angles within the same setting. This consistency is crucial for building a believable and immersive film world.

Animation is performed in Cinema Studio using its multi-shot feature, which allows stringing together multiple shots with different angles and camera movements within a single scene. Character and prop sheets are uploaded as references, with each element named and described.

For prompting, a technique is used where the image is described first, followed by the reference. For example, 'policeman in uniform reference my character enters the locker room with blue lockers reference the location.' This method anchors the AI model to the desired description before pulling from the reference, leading to noticeably better results. Handheld camera movements are often selected to add a realistic, raw feel to certain scenes, enhancing the emotional impact.

Dialogue scenes are crafted to reveal character and advance the plot, often using a single continuous shot to capture natural interactions. For instance, an exchange between the two cops in the locker room is presented without cuts, highlighting their dynamic. To break up dialogue and add visual interest, B-roll shots are generated.

Claude is utilized to create detailed prompts for B-roll, such as 'dash cam POV of a suburban street' or 'extreme close-up of a police radio mic.' These B-roll images are then generated in Soul Cinema and can be mixed into the dialogue sequences, or even overlapped with speech, to enhance the scene's pacing and visual storytelling. This approach allows for dynamic editing and adds depth to the narrative.

The final scene introduces a new character, the protagonist's wife, generated in Soul Cinema without a Soul ID to ensure a grounded and present feel. The scene builds suspense by transitioning from a police call-out to a surprise birthday party. A multi-shot sequence is generated, starting with the cops in the car, then exiting to an outside view, and finally entering the building.

The location for the party, a dark hallway with a closed door and bright light leaking through, is generated with Claude's help. The climax involves the protagonist entering the house like he's on a mission, only for the lights to switch on, revealing a surprise party with his wife holding a cake. This moment provides a touching resolution, showing the character's emotional breakthrough.

All generated shots are brought into Da Vinci Resolve for editing, where shots are trimmed, cut, or mirrored for optimal composition. A key advantage of this AI workflow is the ability to generate missing shots during post-production using the exact same workflow, maintaining consistency in lighting and camera angles without the complexities of traditional filming.

For voice synthesis, audio tracks are split for each character. The protagonist's voice is processed using Hicksfield Audio's 'change voice' feature, applying his own voice while preserving the original emotions and pacing. For other characters, preset voices (like 'Dave') are selected. This final step integrates picture and sound, completing the AI-generated short film.