How it works

A production pipeline, not a prompt box.

Silver is the production pipeline built on your Brand DNA: video, static ads, product photography, image, and motion graphics, one system from research to render. Video is the lead and most-built leg, and the one this page walks end to end: brand and category research, hooks scored, scripts ranked through a compliance gate, then a storyboard. The honest line sits where it belongs, the research-to-storyboard pipeline runs today, and the final render stage is in build. Everything grounds in the same verified brand record as every Jinn product: how the record works.

Apply for early access

Silver is in private beta. Nothing shown on this page is a customer, and no video on this site is a Silver render. The pipeline described below is real and running; self-serve access is landing.

At a glance
What runs today

The upstream pipeline: brand and category research, stored ideation briefs, hooks scored and ranked, scripts written through a four-stage compliance gate, and a storyboard. All of it runs in our production pipeline now.

What it grounds in

Your brand, and every surface reads it. The same verified record grounds video, static ads, product photography, image, and motion; video is the leg furthest along and the one detailed here. It studies your record and your catalog the way a new hire would, then pattern-matches against a library of analyzed in-category videos so scripts are grounded in what actually lands, not a template.

What is in build

The final render, QA, and assembly stage. The architecture is written and described below, honestly, as engineering. It has not run live, so this page describes how it is built, not a result.

What it refuses

To skip the human. The fourth compliance stage is an explicit person’s sign-off, and nothing reaches production automatically. It also refuses to name a render model that is not wired.

The pipeline

Research first. Render last.

This is Silver’s video leg, its most-built path. Five stages that run today, in order, each one persisted so the work is traceable rather than a single opaque prompt. The render stage that follows them is in build; it is described further down as architecture, not as a step you can run yet.

  1. ResearchIt studies your brand from your own record and catalog, then mines a library of analyzed in-category videos for the hooks, beats, and patterns that land in your niche. Scripts are grounded in real analysis, not a template. This stage runs today.
  2. IdeateThe research becomes ideation briefs, and every brief is stored with the inputs that produced it. Traceability is a property of the construction, not a promise: you can always see what a brief was built from. This stage runs today.
  3. HooksEach brief produces a slate of hooks, and a dedicated scoring and ranking agent grades them on strength, clarity, brand fit, novelty, and platform fit. Only the best advance to full scripts. This stage runs today.
  4. ScriptWinning hooks become scripts, each passing a four-stage compliance gate before anything is ranked: a regulatory audit, brand guardrails and voice, platform policy, and a human sign-off. A ranking agent then orders the survivors with the rationale shown. This stage runs today.
  5. StoryboardThe chosen script becomes a shot-by-shot storyboard with the character and product references pinned, ready for the render stage to pick up. This stage runs today; what follows it is in build.

The render, QA, and final-assembly stage that turns a storyboard into a finished video is in build and has not run live. Its architecture is described below as engineering, and it is kept deliberately separate from the stages above, which do run today.

Grounding

Grounded in real analysis, versioned like software.

The scripts are not template fills. Every niche is pattern-mined from real scraped video analysis, so a hook advances because the pattern has landed in your category, not because it sounded good in a prompt.

The prompts that drive the pipeline are versioned and traceable rather than hardcoded, and every ideation brief is stored with its inputs. When a script exists, you can trace it back through the brief, the hooks, and the research that produced it. That traceability is built in, not bolted on.

Compliance

Compliance is a gate, and a person closes it.

Every script earns its way through four stages before it can be ranked, and the last stage is a person putting their name to it. This runs today.

Regulatory auditThe script is checked against category rules first, before brand or platform concerns, so a regulated claim is caught at the top of the funnel rather than after a render.
Brand guardrails and voiceIt is held against your brand’s voice and guardrails, drawn from the same record the research stage studied, so the script sounds like you and stays inside your lines.
Platform policyThen against the target platform’s content policy, because a script that a platform will not run is not a finished script.
A human signs offThe fourth stage is an explicit person’s decision. Nothing reaches production without a human approving it. The gate is not a suggestion the model can talk its way past.
Apply for early access
In build

How the render stage is architected.

This stage is in build and has not run live. Read the rest of this section as engineering, the shape of how the render is designed, not as a result you can produce today.

  • One call, one sequenceThe design produces a coherent multi-shot sequence from a single model call rather than stitching separate clips together, so the shots hold together instead of cutting between mismatched takes.
  • Audio is model-nativeAudio and lip-sync are generated in-model, not bolted on by a separate text-to-speech stage, so the voice belongs to the shot instead of being pasted over it.
  • Identity is anchoredCharacter identity is anchored by multiple reference images, and the product reference is injected only into the shots that actually show the product, so it does not bleed into frames it does not belong in.
  • Spend is reserved firstBefore any paid render call, spend is reserved against a per-brand budget, so a run cannot quietly overspend. A designed vision-judge and budgeted regeneration loop sits above it, and the page is honest that both are in build.

The quality judge in this design grades a single decodable frame, not full motion, and its thresholds are spike-grade ports awaiting a live render. It does not gate anything today, and the page will not claim it does.

Honest machinery

Honest about what runs and what does not.

Nothing shown is a customerSilver is private beta. No output on this site is a Silver render, and there is no claim of a live, self-serve, or customer-used product. Jinn is its own first customer while the render stage lands.
A human is in the loopThe compliance gate ends in a person’s sign-off, and nothing reaches production automatically. The human is a required stage, not an optional review.
The work is traceableEvery ideation brief is stored with its inputs and the driving prompts are versioned, so a script can be traced back through the brief and the research that made it. Traceability is built into the pipeline.
The record underneathThe pipeline grounds in the same verified brand record that powers every Jinn product: 346 fields structured from your own site and evidence. Silver reads the record. It does not invent one.
The substrate

Silver sits on the record.

The brand record the whole pipeline grounds in is shared substrate, documented once and linked here rather than re-explained on every product page. The terms this page leans on are defined in the dictionary.

How the record worksThe research programVoice check
Questions

The questions people ask.

Can I buy this today?
Not self-serve yet. Silver is private beta: the research-to-storyboard pipeline is real and running, and the render stage is in build. Join the waitlist and we will bring you in as it lands.
Is anything on this page a Silver video?
No. No output on this site is a Silver render. The page describes the mechanism, and it draws an honest line between the stages that run today and the render stage that is still in build.
Is it fully automated, or does a human review?
A human approves before production. The fourth compliance stage is explicitly a person’s sign-off, and nothing reaches production automatically.
How does it keep my brand voice consistent?
It studies your brand from your own site and catalog first, then enforces voice guardrails as one of the four compliance stages before a script is ever ranked, all grounded in the same verified record.
Why does the render stage get such a careful caveat?
Because it has not run live yet, and we would rather show you the architecture honestly than imply a finished capability we have not proven. The upstream pipeline runs today; the render stage is described as engineering, not a result.

The pipeline is real. Get in line for the render.

Research, hooks, scripts, and compliance run today. Join the waitlist and we will bring you in as self-serve and the final render land.

Apply for early access