avatar-explainer-videolisted

# Avatar Explainer Video (Higgsfield MCP) ## Overview Turn source content (PDF, doc, article, topic) into a short cinematic explainer video: a realistic AI presenter — optionally the user themselves — explains the content on camera, with cutaway scenes visualizing the key ideas. Everything runs through the **Higgsfield MCP** (no CLI, no local ffmpeg needed). **Core principle: spend words before credits.** Read, script, and plan for free; generate once, well; ask the user for approval only *after* a scene is generated — never before. **Requirements:** Higgsfield MCP connected, a paid Higgsfield plan with credits. Check credits early with `balance` / `show_plans_and_credits`. ## The Flow ``` 1. Read source → extract 3–6 key points → draft script 2. ONE interview round (questions below) 3. Identity setup (photo → Element) [free-ish] 4. Anchor still → show it → user approves [~1 credit] 5. Scene videos, one at a time → show each → user approves [~30 cr/scene] 6. Assemble with explainer_video [free] 7. Deliver + optional upscale ``` ## Step 1 — Read the source Read the PDF/document the user provides. Extract: - The 3–6 ideas that matter (an explainer is not a summary — pick what a viewer must understand) - A natural narrative arc: hook → mechanism/how-it-works → payoff/so-what Draft the script now (see Script rules) so the interview can show a concrete plan. ## Step 2 — The interview (ONE round of questions, then stop asking) Ask all of these in