VN-DEMO-004: allow adult-specific ElevenLabs voice override

VN-DEMO-004: process first four adult narration gaps

VN-DEMO-004: fix adult manifest cost fields and empty descriptions

VN-DEMO-004: widen adult direct video inputs and correct cost estimates

VN-DEMO-004: add adult demo spend guardrails

VN-DEMO-004: switch adult demo to enabled Replicate LLaVA fallback

chore: redeploy after enabling Vercel Web Analytics

VN-DEMO-004: adult.vnpoverview.com - LLaVA-1.6 + zero-data demo

VN-DEMO-003: anonymous run analytics via Vercel Analytics

Log Cobalt fallback failures

Normalize YouTube playlist URLs before download

Add Cobalt YouTube fallback and download hint

fix: add EXPOSE 8000 for Railway HTTP routing

chore: force Railway redeploy trigger

fix: run uvicorn from /app/demo-api to avoid hyphen in module path

Remove startCommand from railway.toml — Dockerfile CMD handles PORT expansion

Switch to Dockerfile builder — Nixpacks incompatible with non-standard layout

Fix nixpacks: use standalone pip binary (not python3.11 -m pip)

Fix nixpacks: add python311 to setup phase, use python3.11 -m pip

Add demo app — FastAPI SSE backend + Next.js frontend (VN-DEMO-001)

Add vn score — GPT-4o-as-judge AD quality scorer (VN-013)

Add vn podcast — FFmpeg audio mixing podcast generator (VN-012)

Add vn ad — WCAG Standard AD track generator (VN-011)

Add vn sports — Rekognition + IoU tracker + GPT-4o sports narration (VN-010)

Add vn theater — ElevenLabs TTS audio description track (VN-009)

feat(VN-008): Educational Video Describer — education-specific GPT-4o prompt with NO_VISUAL_AID filtering

feat(VN-007): Festival Film Accessibility Kit — gap-targeted GPT-4o narration

refactor(gaps): replace Whisper with Deepgram Nova-3 + FFmpeg silencedetect

feat(VN-006): Compliance Check Report — WCAG/CVAA gap coverage scoring

feat(VN-005): Narration Gap Finder — Whisper dialogue gap detection

feat(VN-004): Narration CLI — pip-installable vn command

chore: remove session handoff doc post-merge

Implement VN-003 Rekognition object detection

Add API server, engine modules, Lambda handler, and .gitignore

Add live API demo results (January 2026)

Update links: remove broken docs URL, point to HF Space

Merge: keep YAML metadata in README

Update model card with YAML metadata and accurate claims

Update with verified claims

security: Remove API keys and add environment template

docs: Sync README with Hugging Face version and add demo links

feat: Add phase3 and phase6 directories

docs: Update README with clear demo separation and HF links

fix: Use valid Hugging Face dataset IDs in YAML metadata

docs: Add proper YAML metadata to model card

feat: Visual Narrator 3B - Clean repository with professional benchmarks