How I Made an Animated Video Using AI Only

The Soul Publishing supplied the script. I used ChatGPT to break it into scenes, character descriptions, props, lighting and transitions, then developed reusable prompt structures. These early concepts explore the cast and the atmosphere before the visual direction settles.

Image development included experiments with Midjourney, Google ImageFX, Leonardo AI, Ideogram and Helio AI. Rooms, props and light establish the setting; recurring environmental cues make separate generated images feel connected.

Locked character descriptions, wardrobe details and lighting cues helped reduce visual drift. Expressions, gestures and camera distance give the characters a point of view, while repeated details support continuity across the scene variations.

The mechanical lock became a practical test of AI’s limits. Generated gears and letter dials did not always align or behave convincingly. Photoshop refinement brought the details closer to the intended design, keeping the object readable within the story.

Warm light, balloons and group compositions shift the emotional register. Reaction shots and birthday details contrast with the darker doorway scenes, giving the visual sequence a clearer change in mood.

Image-to-video experiments included Runway, Pica, Sora and Helio Video AI. These supplied files preserve individual story frames; they are still images rather than playable motion clips. The focus is continuity between the source visuals and the intended sequence.

Manual image refinement continued in Photoshop. The Soul Publishing supplied narration created with ElevenLabs; I shaped the wider soundscape with Envato Elements effects, including room atmosphere, doors and party sounds. Premiere Pro brought the scenes, audio and text together. The timelines show the layered assembly behind the result.

The original project cover closes the collection. This experiment demonstrates an AI-led workflow supported by creative judgment, manual refinement and editing throughout.