I began with free time and a question: how far could I take an idea with the AI video tools available then? I did not expect a few minutes of film to demand such sustained direction, selection and editing.
Making ten seconds belong to the same film.
The hardest part was keeping a world coherent across generations. Each scene was only around ten seconds, but its colour, atmosphere, characters and action had to belong to the same film. I built a small prompt-development app inside Google Omni to prepare scene prompts against that vision and keep continuity in view.
A toolkit, shaped around the idea.
No single tool handled the whole process well, so I used different tools for different jobs. Nano Banana helped with storyboards; ChatGPT with the script; ElevenLabs with voiceover; Omni and Veo with video generation and sequencing; and CapCut with the final edit. I directed the choices, rejected generations that did not fit, and shaped the sequence and pacing into the finished cut.
Nano Banana · storyboards · ChatGPT · script · ElevenLabs · voiceover · Omni + Veo · video · CapCut · edit