The platform underneath

Built on AWS, so ten clips take about as long as one

Enrichment and rendering fan out: every clip gets its own worker, and every composition renders on its own serverless invocation. A busy hour costs you wall-clock, not a bigger machine and not a queue.

☁️

Elastic where it matters

Each enrichment run is dispatched through a queue and executed as its own container task; each composition renders serverlessly. Ten clips run ten at a time rather than one after another — the difference between a weekend of sport being a bottleneck and being a Saturday.

🔒

The team is a hard boundary

Assets, packages, agency credentials, AI keys, delivery targets and feeds are all scoped to a team, with admin and member roles, invitations you can cancel and access you can revoke. Sign-in is Google OAuth; sessions are short-lived and every API route is authenticated.

💬

It tells you what happened

Slack notifications for library activity, delivery activity, alerts and new material in a feed — each to a channel you choose. A burst collapses into one message instead of a hundred. You can stage a story straight from the notification.

Storage, media and money leave a trail

Assets live in object storage, with server access logging and cloud audit trails on the data events, both on a retention schedule.
Credentials are redacted in logs as a deliberate pass, not a hope.
AI spend is recorded per run — provider, model, units, and which account was billed — and surfaced in Settings rather than arriving as a surprise invoice.
Long jobs get proportionate timeouts, scaled to media duration rather than one flat number that is wrong at both ends.
Every delivery keeps a timed step log, so "it failed" is always answerable with "at which step".

The shape of itAWS

API & databasecontainers + managed Postgres

Media & assetsobject storage

Enrichment workersqueue + container task per job

Composition renderingserverless

Infrastructuredefined in Terraform

Where it is honest to be modest. The fan-out is real and it is the part that matters for throughput. The API itself is not yet horizontally scaled — it has not needed to be. If you are sizing this for a large roster, ask, and we will tell you exactly what would have to change.

Help that knows where you are

An operator's guide, in the product

Eleven topics covering every screen, reachable from the screen itself.

Contextual

A ? beside each screen title opens the topic covering that screen, and each clip carries help for the step it is on.

Answers, not improvisation

Ask MediaBlaze answers only from the shipped guide. It cannot act, and it does not invent a feature that isn't there.

Written like this site

The same rule applies in the guide as here: name the concrete outcome, state the limits, and never describe a step the product doesn't take.