Atmosphere as a Service. Infinite, on-brand, license-free music — generated per moment, per zone, per activity. Pre-revenue v1 with the architecture already done.
Brand mood → channel → zone → schedule → generated track. The buyer skips a half-year of engineering and starts on day one with the loop already closed.
Every café, gym, hotel lobby, retail store and restaurant owes PRS, PPL, ASCAP, BMI, SACEM. A mid-sized UK venue: £500–£2,000+ per year just to play recorded music in earshot of customers. Globally — multi-billion.
Spotify's "coffee shop" sounds the same in a Shoreditch roaster as in a motorway service station. A 6am spin class and a 7pm cooldown yoga get the same beats. Brand cannot live in someone else's catalogue.
Soundtrack Your Brand, Mood Media, Cloud Cover: £25–£60 per zone per month, plus passthrough licensing. Royalty-free libraries: the same 4,000 tracks every other venue is also using. All three: pre-recorded, finite, generic.
Mureka, Suno, Udio, Stable Audio now produce commercial-quality music on demand from a text prompt. Every constraint above inverts.
AI-generated music with no human authorship sits outside the collecting-society repertoire. PRS for Music has stated this publicly.
Every track is new. No repetition fatigue. No "we heard this one yesterday" from the regulars.
The prompt encodes the venue's identity, time of day, zone activity, energy of the moment. Sound becomes a brand surface.
Lobby gets one thing, gym floor another, both shift through the day automatically. No DJ, no playlist curation.
Not a wireframe, not a stub. A working v1 with the architecture done right.
ADR-081 codifies tenant isolation by venueId on every tenant-scoped query. JWT via jose, argon2 password hashing. Cross-tenant features require an amending ADR — the boundary is not accidental.
db for testabilitydb last (ADR-082)AppError for domain failuresFive venue-level sliders define sonic identity. Channels (how music sounds) decouple from Zones (where it plays). A chain with twenty physical rooms draws from a shared brand palette without copy-pasting config.
"Spin class Mon/Wed/Fri 6–7am in Studio A, with this energy profile." A react-big-calendar + @dnd-kit interface that drag-drops the venue's week into existence.
ADR-068 defines the seams. ADR-069 makes prompt composition a single typed boundary — two entry points, Zod-validated at the Mureka edge. Swap Mureka for Suno or your own model without touching the rest of the system.
One device per zone, enforced with heartbeat. Server-sent events push now-playing and a remote-kick signal to playback devices. The venue's tablets just work.
A full set of billing decisions lives in .docs/project/prds/account-billing/. Implementation pending — the buyer is wiring up a model that has been grilled, not inventing one.
Single per-minute rate in 5-minute increments. Rate bundles generation, delivery, sharing-pool credits, repeat-playback. No tiers, no platform subscription line.
Thin BillingPort interface. Paddle adapter first (Merchant-of-Record handles tax). Operator chooses adapter at deploy time, not the venue at checkout.
Hold at job enqueue, capture at delivered cost, release on terminal failure. Mureka retries inside the job are free — provider charges only on successful delivery.
Generation halts immediately; playback loops existing inventory through the end of today's last scheduled session in venue-local time; hard stop after.
Scheduled session live, no registered + logged-in + actively-playing + volume-up device present — that's a violation. Randomised server-driven heartbeats. Attribution to session and assigned staff.
Active → paused → restartable → deleted-on-request. Dormancy auto-pause after 7 days. Free-trial uses the shared-song pool to minimise generation cost.
You already run at least one SaaS. You can stand up Redis, point a domain, wire Paddle, and sign a pilot deal with a regional gym chain or coffee group without needing a roadmap from me.
You see the licensing-tax angle and the brand-expression angle. You want the engineering already done so you can spend year one on go-to-market instead of architecture.
If that is you, the technical risk is removed. The remaining risk is the one you are good at.
SideProjectors / MicroAcquire / Acquire.com bands, with Venue Beats positioned on architectural grounds.
Venue Beats sits in the upper pre-revenue band because: a sharp non-obvious market wedge (PRS/PPL replacement), architecture genuinely above MVP quality, multi-zone scheduling + device presence + SSE streaming is 3–6 months of engineering the buyer skips, and the agentic development framework is differentiated on these marketplaces — most listings ship code, not a way to keep building it.
PRS for Music has publicly stated that purely AI-generated music with no human authorship is not eligible for copyright and therefore falls outside their licensing repertoire. A formal barrister's note codifying this position for commercial venue use is the next priority.
Small VPS, public click-through URL — buyers experience the scheduling UI and a generated track playing in a zone.
Brand-mood → channel → zone → schedule, ending in a generated track played live.
Wiring up the already-designed BillingPort — prepaid wallet, auto-top-up, hold/capture on generation jobs, dunning, invoices. (The model is locked; this is implementation, not design.)
Any signed letter of intent, paid or unpaid, shifts the asset from "code" to "code with first customer."
Even 50 emails reframes this as a warm asset rather than a cold codebase.
For the buyer who needs to scale beyond single-server SQLite.
You get the full v1 codebase, the agentic development framework, and the architectural head-start. You take on the commercial risk that the roadmap above is designed to remove.
As each roadmap item lands, the price moves up. The legal opinion alone reframes the entire commercial thesis. The first signed LOI reframes the asset class.
Working title only. No public deployment, no domain in use, no logo, no customers — nothing branded. The buyer renames it on day one.