---
title: A family tree that plays
summary: Work out the family from a few recorded relationships, draw it as an animated tree, and give each person a chapter of their best footage.
lede: Every branch gets a chapter of that person's best footage
outcome: people
tools: family_of, get_person, person_locations, quick_cut_plan, reframe_person, list_voices, synthesize_speech, render_html, render_concat, mix_audio, render_ffmpeg, job_wait, create_project, commit_to_library
works_with: claude-design
order: 0.7
---

> Work out the whole family from p88 (Nichole) in GroundSlate and give me everyone in it with their name, how many clips they are in and the cities they were filmed in. Have Claude Design draw that as an animated family tree at 1920x1080, generations fanning in oldest first, lines drawing between them. Then give each person a chapter: six seconds of their best footage with the crop following their face, only one clip per day so nobody swamps their own branch, playing beneath the tree while their branch lights up. Nothing on screen but the tree, the names and the footage. Narrate the names in Nova's voice, the one attached to avatar av_4d5e6f7a, duck the bed at ~/Music/beds/quiet.mp3 underneath as one continuous mix, and file it in a personal project called "The Family". Tell me anyone who got no chapter because they were never recognised in their footage, and which chapters lost the face for part of the six seconds.

- Record two links, a daughter and a son, and the siblings, cousins, aunts and grandchildren are worked out from there.
- `reframe_person` keeps each person's face in frame: one call per chapter, naming the person. The planner picks the moment; the reframe renders it.
- Faces must be indexed first; face recognition is one of the steps in the app's setup. A clip where that person was never recognised is refused rather than centre cropped, so it cannot become their chapter.
- When a recognised face drops out for a moment, the crop holds, then drifts to the shot's main subject, or the centre, until they are back. Each chapter reports how much of it stayed on the face, which is what the last line of the ask reads back.
- The narration is spoken on your Mac in the avatar's voice. No provider or key needed.
- Ask for one continuous mix, or it plays as a stack of clips with different room tone.
- Each chapter is its own short render, joined into one film.