One client brief in. A complete multi-week, multi-channel campaign out.
Paladia turns a single brief into a whole campaign arc — copy, image prompts, video scripts and email sequences, written for each platform rather than copy-pasted across them. Nine platforms, four campaign weeks, one shared campaign thread that every asset inherits.
Built by 92eleven as a white-label engine an agency can put its own name on. This build runs end to end with no API keys at all.
Read this as a demo, and read the numbers as engine output
Every string below was produced by the engine running in mock mode with zero API keys. No copy is hand-written for the site, no performance figure is invented, and no live provider, publishing or billing call exists in this build.
Mode: mock · live mode enabled: false · connectors configured: 0/9. Copy generation is deterministic templating, not an LLM call; image prompts and video scripts are text only — nothing is rendered, published or billed.
Four deliverables per week, per platform
Platform-native copy
Headline, body and CTA sized to each platform's ceiling — 280 characters on X, 3,000 on LinkedIn, no hashtags on Pinterest, an affiliation disclosure on Reddit.
9 platform variants per week
Image prompts
A direction-ready prompt with palette, light, framing and aspect ratio, a negative prompt and alt text for accessibility.
prompt text only — no image is generated
Video scripts
Timed beats from hook to CTA with direction per beat, on-screen text, a shot list and audio notes.
script and shot list only — nothing is rendered
Email sequence
One email per campaign week: subject, preheader, body, CTA, segmentation rule and a required unsubscribe line.
written to JSON — nothing is sent
Six agents, one shared campaign thread
Maps the brief into the week-by-week arc, assigns each week a role and writes the shared thread (hook / angle / pillar + thread lock).
seq 1 · 0 ms
Writes the headline, body, CTA and the email entry for every week.
seq 2 · 0 ms
Adapts that week's copy for all nine platforms in each platform's own register and length ceiling.
seq 3 · 0 ms
Validates every asset against the brand DNA, asserts zero thread drift and reports brand health.
seq 4 · 0 ms
Attaches labelled attribution placeholders — bands and policy fields, never a number.
seq 5 · 0 ms
Assembles the payload the dashboard renders, and refuses any asset that reached it without a quality result.
seq 6 · 0 ms
Week 3 on LinkedIn and week 3 on Instagram carry the same hook and angle
Our most useful feedback came from people who never left a review.
Let repetition be the testimonial — what people reach for without thinking.
Worn In, Not Bought
Shown from Full Frequency Co.'s campaign — the same three fields appear on all 9 assets in this week.
Our most useful feedback came from people who never left a review.
Our most useful feedback came from people who never left a review. Honestly, here's the plain version, no wrapper.
Our most useful feedback came from people who never left a review.
Our most useful feedback came from people who never left a review. Honestly, here's the plain version, no wrapper.
Different body length, different hashtags, different register — identical hook, angle and pillar.
No client's data can reach another's
Full Frequency Co.
Lifestyle Appareltone: Warm & Directpeople who dress for the day they actually have, not the one in the lookbook
Torczon Digital
Graphic Appareltone: Bold & Playfulcollectors who buy the print, not the logo, and know the run size before they ask
A quality gate and a spend gate
Quality Gate
PASS · pass rate 100%Every asset is scored 0–100 (Brand Health Score) against brand voice, platform format, thread consistency and completeness. Hard failures — a fabricated metric, copy over the ceiling, a thread mismatch — fail the asset outright and route it back through the creative agent exactly once.
Two-tier spend gate
kill-switch disarmedTier 1 is a soft limit at 70% of the campaign budget — the action is allowed but flagged for a human. Tier 2 is the hard limit at 100%: the action is refused and the kill-switch engages, halting every autonomous financial action until a named human clears it by hand.
Drop your own keys in and the switch flips
if (mode !== "mock" && live_mode_enabled === true) {
// a provider call MAY be attempted
} else {
// deterministic mock generators run;
// no outbound call is made, ever
}In this build mode = "mock" and live_mode_enabled = false, so no connector is reachable at all. Nothing here publishes, sends, renders or bills — the engine only writes JSON.
Everything on this site is engine output.
72 assets across 2 clients and 9 platforms, scored by the gate, with the spend gate's ledger and kill-switch state included. Open the Command Center to read the pipeline log, or jump straight into an asset.