What you're looking at.

Casset is a production-grade audiovisual music identity platform and artist-home layer for the generative media era. It is built around Profile Worlds, Hook Objects, Release Rituals, and Listening Rooms: the layer where a song becomes emotionally legible before the rest of the internet flattens it into a streamable file, feed item, or machine asset. Underneath that surface, Casset is adding release manifests, provenance, permissions, contributor context, lineage, and agent-readable access routes.

What you have

A Next.js App Router app, a deep Prisma schema, a custom cross-browser audio engine, Visual Studio/runtime systems, Stripe Connect commerce, an SSE-based realtime layer, co-cassets, and a complete onboarding-to-room loop. Today, from a listener's perspective, that's:

• Paste casset.fm/yourname into iMessage → branded preview card.
• Tap it → the Hook Object plays instantly, no streaming app in between.
• See the world around the song: atmosphere, profile identity, provenance, and room state.
• Follow, join, collect, reply, or share → the release gains memory.
• Tap Share → vertical audiovisual artifact exported for social surfaces.
• Internally, the release can resolve to manifest, permission, provenance, anchor, and lineage state.

Tech stack

• Next.js 15 App Router, React Server Components, TypeScript strict.
• Postgres via Prisma; Vercel Blob for audio + artwork.
• Stripe Connect (Destination) for payments; Apple Pay / Google Pay first.
• Redis for presence, rate limits, short-term caches.
• Vercel Edge Middleware for auth + redirects.
• Sentry + Vercel Analytics; vitest for tests.

The core systems

1. 1
   Audiovisual playback engine
   A bespoke cross-browser layer enforcing 30s hook windows with one playback clock for audio, lyric timing, waveform state, shader treatment, and mobile-safe rendered artifacts. The hook resolves into one playback surface rather than separate audio, lyric, and visualizer layers.
2. 2
   Audio access + security
   The raw audio URL never leaves the server. Unentitled listeners get byte-range-clipped proxied audio; entitled listeners get minute-lived signed URLs. Entitlement is re-resolved on every stream.
3. 3
   Commerce
   Stripe Connect charges, Apple Pay / Google Pay as the default confirmation, webhook-authoritative grants, referral attribution, per-track rewards, campaign payouts, refund-aware entitlement. See commerce.
4. 4
   Release quests / drops
   Time-boxed prize pools with a scoring pipeline that blends views, engagement, and integrity signals. Important as optional Release Ritual infrastructure, not as the platform category.
5. 5
   Listening Room / Side B
   Every casset has a room layer: SSE-streamed activity feed, live presence indicator, comment threads (on the casset + per-hook), custom paid emoji, follows, and tip support.
6. 6
   Hook sharing
   Visual Studio creates exportable audiovisual hook identities: hook audio, artwork, shader atmosphere, lyric treatment, waveform, and CTA can resolve into vertical cinematic MP4 artifacts for mobile sharing.
7. 7
   Artist Studio
   Inline editable cassets, preview scrubber with pre-computed waveforms, Visual Studio, price editor, drop manager, clip-submission review, Stripe onboarding, cover/theme controls, profile textures, and AI/No-AI provenance badges.
8. 8
   Realtime
   SSE over HTTP/2 instead of WebSockets — cheaper on serverless, survives middleboxes, powers comments, activity feed, and presence from a single long-lived endpoint.
9. 9
   Provenance and identity controls
   Creators can make authorship legible through profile details, verification state, AI profile and No AI used profile badges, profile textures, visual identity, social links, and controlled theme state.
10. 10
    Canonical release layer
    Release, ReleaseVersion, ReleaseManifest, ReleaseAnchor, Contributor, Split, PermissionPolicy, ProvenanceEvent, DerivativeLink, and AgentAccessPolicy are implemented as the hidden dossier layer under the cinematic product. Agents can read manifests, permissions, provenance, lineage, license routes, and access decisions for published canonical releases.

What makes this hard to reproduce

Cross-browser audio is an iceberg

Every browser handles byte-range requests, MediaRecorder, autoplay policies, and AudioContext resumption differently. Casset ships consistent playback + export across Safari, Chrome, Firefox, iOS PWA, Android PWA, and in-app browsers (TikTok/Instagram) — that alone is months of quiet work.

Stripe Connect is not a weekend integration

Destination charges + connected-account onboarding + webhook-driven entitlement + refund + dispute handling + referral splits is months of compliance work. Casset has all of it and the dead-letter handling to survive a webhook retry storm.

The campaign engine is fintech-grade, but strategically secondary

Campaign payouts are real money. The engine records scored snapshots over time, applies integrity penalties, writes payouts through an append-only ledger, and releases via Stripe Connect transfers. It is a credible support system for release quests, but it should not be the first thing investors think Casset is.

The AI posture is provenance-aware

Drop strategy recommendations are computed over historic engagement snapshots, not freeform LLM calls. Profile badges make AI/no-AI identity explicit without moralizing. The product treats AI as creative support, not the headline category.

Generative abundance makes identity and permission more valuable

As generated sound and visuals multiply, durable value shifts toward artist identity, release context, consent, provenance, and permission routes. Casset is positioned as the home for that context, not as the system generating the most content.

Architecture overview

See the architecture doc for the full system picture. The short version:

• Browser (PWA) — talks to Next.js for HTML/RSC/JSON and to the SSE endpoint for realtime.
• Next.js — routes, middleware auth, API. Static for marketing; dynamic SSR for identity; client islands for playback.
• Postgres — Prisma-managed, migrations via prisma migrate deploy only in production.
• Vercel Blob — private-read bucket keyed by content hash for audio + art.
• Stripe — Connect + Apple Pay + webhook-source-of-truth.
• Redis — presence, rate limits, short-term caches.

Data model (high-level)

• Identity & content — User, OAuthAccount, Artist, Track, ArtistMedia.
• Community & activity — CassetActivity, ActivityEvent, CassetComment, HookComment, HookLike, Follow, Emoji, TrackReaction.
• Campaigns & rewards — Campaign, CampaignParticipant, CampaignSubmission, SubmissionMetrics, SubmissionScore, CampaignPayout, Bounty, HookVideoSubmission.
• Intelligence & reputation — DropIntelligenceSnapshot, DropStrategyRecommendation, PromoterProfile, PromoterReputation.
• Accounting — CreditLedger (append-only), Purchase, ReferralAttribution, PerformancePayout.
• Canonical releases — Release, ReleaseVersion, ReleaseManifest, ReleaseAnchor, Contributor, Split, PermissionPolicy, AgentAccessPolicy, ProvenanceEvent, DerivativeLink.

Build estimate

A realistic backfill estimate for a small team:

• 1 senior eng, serial: ~30 weeks.
• 3 engineers + 1 designer, parallelized: ~12–16 weeks.
• Dollar-equivalent at US rates: roughly $250K–$400K.
• Critical path: audio engine + commerce are non-negotiable gates.

Honest assessment

What's production-ready

• Audio pipeline + Hook Object playback + profile media footer.
• Commerce (Apple Pay + webhook-authoritative grants).
• Artist Studio, preview scrubber, Visual Studio, profile textures, provenance badges, rewards toggles.
• Listening Room / Side B (comments, presence, activity, follows) and co-casset rooms.
• Canonical release models, manifests, agent-readable permissions, provenance, lineage, and Base anchoring workers.
• OG image rendering + TikTok export.

What's built but may need iteration

• Campaign scoring + payouts (live, but edge cases keep surfacing).
• Drop advisor (ships recommendations; UI surface still evolving).
• Promoter reputation (graph is built; surfacing it is ongoing).

What doesn't exist yet

• Label / multi-artist management in one account.
• Public analytics for fans beyond the activity feed.
• Public self-serve licensing UI for artists and agents.
• Native mobile wrapper (PWA is the shipped mobile surface).

Why this is hard to copy

• Audiovisual identity shape. The whole product assumes the hook is an audiovisual object inside a Profile World — not a track, not a playlist. Retrofitting a generic music link tool to match means rebuilding the schema.
• Stripe Connect is a moat of its own. Onboarding, compliance, and webhook discipline take quarters, not weeks.
• Room and ritual memory. Comments, follows, clips, co-cassets, collects, presaves, and release-room traces become a social memory graph around music.
• Release context for the agent era. Manifest hashes, contributor graphs, permission policies, provenance events, and lineage are expensive to retrofit after a product has trained users to think only in tracks and links.

Recommended next steps

If hiring engineers

Start with one senior full-stack familiar with Stripe Connect or streaming audio. The codebase rewards craft — optimize for that rather than for output velocity.

Quick wins (<1 week each)

• Make Profile World and Hook Object language consistent across the app.
• Surface provenance, Release DNA, and room activity more clearly in Studio.
• Add ritual/room summaries to the creator dashboard.
• Turn on push notifications for iOS 16.4+ PWA installs.

Medium-term (1–3 months)

• Multi-artist accounts (label mode).
• Release Ritual dashboards that foreground room memory before payout mechanics.
• Artist-authored permission presets for AI training, AI generation, remix, sync, stems, and derivative registration.
• Native mobile wrapper or App Clip for instant collect flows.