Turn script lines into lipsync avatar videos

The script writing workflow calls this with a script id. The first half loops over the script's lines, finds each line's character, and turns the dialogue into speech with that character's ElevenLabs voice. The audio is uploaded to Cloudinary and saved on the line with the character image.

It then calls its own second webhook. That half loops over the lines again: an OpenAI vision model looks at the character image and the dialogue and writes a two-shot talking avatar prompt. The image, audio and prompt go to the Creatify Aurora model on fal.ai. The workflow checks until each clip is ready and saves it to the line. The reel workflow stitches the clips together.