Create AI characters with image and voice prompts

This is the first step of the character video system. You start it from an Airtable button on a generation config record that holds your theme, tone, setting, character angle and how many characters you want.

If you chose one shared background, an AI agent writes a background prompt and GPT Image 1.5 on fal.ai renders it. Next an agent using a model through OpenRouter writes the cast: a name, a short role, an image prompt and a voice prompt for each. Old characters for that config are removed, the new ones are saved, and a 9:16 video project and a generation log are created. Then it calls the image workflow.