Create AI characters with images and voices
Describe a theme in Supabase and get a cast of AI characters, each with a generated portrait, a designed voice and a video project
Version 1
What it does
Three webhooks run three stages, each using a character config row in Supabase (theme, roles, number of characters, tone, setting, look and voice personality).
The first stage asks a model through OpenRouter to invent the characters, each with a name, role, image prompt and voice prompt, saves them to a characters table (replacing an earlier run) and creates a video project row. If you ask for a shared background, it also makes that image with GPT Image on fal.ai. The second stage makes each character's portrait with GPT Image, on the shared background when there is one. The third designs a voice for each character in ElevenLabs and saves the voice id and sample. A later workflow makes the videos.