Strata

An AI-guided audiobook platform that turns public-domain classics into immersive, narrated experiences — guided commentary, ambient atmospheres, dynamic abridgment, and grounded book chat.

Live at stratabooks.app · POC stage.

Screenshots

Hero / How it works	Library

Why this exists

Audiobooks of classic literature are either robotic TTS or professional narrations with no context. Neither gives you the experience of reading a book with someone who knows it — the pauses where they explain a reference, the atmosphere that matches the scene, the ability to ask "what did that mean?"

Strata builds that. Each episode is a chapter with AI commentary woven in, ambient sound matched to the scene, and a chat that's grounded in the actual text — not hallucinated.

What it does

• Guided narration — chapters narrated with commentary woven in at natural pauses
• Ambient atmospheres — scene-matched sound design (rain, crowds, silence)
• Dynamic abridgment — skip the boring parts without losing the thread
• Grounded book chat — ask questions about the text, get answers backed by passages
• Commentary trust layer — full-text grounding, automated verification, admin review gate, interpretive disclaimer
• Tier-checked audio streaming — the paywall holds on every byte

How it's built

• Next.js 14 (App Router) + TypeScript + Tailwind — full-stack
• PostgreSQL via Drizzle ORM — shared/schema.ts is the source of truth
• Inngest — orchestrates the per-episode pipeline (segment → commentary → verify → abridge → narrate → mix) with pause/resume/retry
• FFmpeg — audio assembly, segment stitching, atmosphere mixing
• ElevenLabs + Microsoft Edge neural voices — TTS driver with provider fallback (free Edge narration, ElevenLabs for flagship content)
• Stripe — webhook-idempotent billing, entitlement read from DB
• Storage-agnostic audio driver — Supabase Storage or Cloudflare R2 in production, local disk in dev

The pipeline is the interesting part. Each episode goes through six stages orchestrated by Inngest, and every stage can fail, retry, or be paused for human review. The commentary layer has a trust pipeline — full-text grounding, automated verification, admin review — so only vetted commentary reaches listeners.

Run it locally

cp .env.example .env.local    # DATABASE_URL, JWT_SECRET, provider keys
npm install
npm run db:push               # sync schema to your database
npm run dev                   # http://localhost:5000

Scripts

Command	What it does
npm run dev	Next dev server
npm run build / npm start	Production build / serve
npm test	Vitest unit/integration tests
npm run lint	ESLint
npm run db:push	Push schema to DB (Drizzle)

CI runs typecheck + lint + tests on every PR.

License

Proprietary. All rights reserved.