Celestia Intelligence
Your audience is already talking.We read it live.
Every run reads what people say and what they do across public platforms, fresh at the moment you ask. Then it builds a marketing plan for your business, or gives you a straight answer on an idea.
What do you want to do?
Domestic-power themes resonate loudly with thriller audiences; the comps are the weak leg and talent momentum is only mixed — refine packaging before greenlight math.
Re-angle it. The appetite is on the class-divide angle: it is the one reading positive and rising, where the other angles are split or upset by the subject. Lead the pitch with it.
The audience read
Audience angles3 scored · 1 unscored (degraded / off-topic)
42 points separate the strongest angle from the weakest, so the shape is a real gap rather than a stretched one.
- 1Class divide and exploitation of trust in domestic workerspositive ↑
- 2Gaslighting and psychological manipulation inside the homeupset by the subject →
- 3Female-led crime and domestic thrillersmixed ↑
Summary
Audience chatter around domestic-worker power dynamics is loud, current, and majority-positivec1c2. The concept's genre lane — the female-led domestic thriller — is rising on both behavioral and conversational signalc5.
The main liability is the comp slate: one of the two comps is a prestige outlier whose audience profile does not transfer to this packagec6c7. Talent momentum is flat rather than falling, which reads as neutral for packagingc8.
For the idea
- The class-divide angle is positive and risingc1c3
- Viewers side with the worker nobody believesc1c3
Against the idea
- The comp slate leans on a prestige outlier, so the financial translation is unreliablec6c7
- A neat ending would lose this audiencec1
What people say. What people do.
One engine behind both ways in: two kinds of evidence, read fresh on every run and read together.
Stated
Posts, comments, reviews, threads, press. Something a person chose to write. It explains the attention.
- YouTube
- X (Twitter)
- Bluesky
- Steam
- Twitch
- GDELT
- Hacker News
Observed
Searches, views, tickets, ratings. Nobody is asked anything; behaviour is simply counted.
- TMDB
- Wikipedia
- Box Office Mojo
- RT / Metacritic
- Google Trends
- Google Ngrams
The four grades
Say or do
1Say
Do3
- Audience resonanceposts, comments, reviews
- Genre fitnesssearches, page views
- Talent momentumpage views
- Comp evidencebox office, ratings
Every report shows what each grade was read from.
Pick a business.Get the plan and the posts.
Nothing to type. The run starts from your business profile, reads everyone talking about your customers' worries right now, and writes a marketing strategy and the posts to make for your channels.
New run
Businessthe one this plan is for
Ledgerly (sample)
2 customer groups · 1 channel
Focusthe whole business or one product
Channelswhere the posts go
Ready
Three picks. No typing.
Your business, the whole of it or one product, and the channels the posts are for.
Your customers aim it. They never grade it.
Their worries pick the subjects. We read everyone talking about them, never your brand, and say which angles reach them.
A strategy, and the posts to make.
Who to reach, where they talk, what to say in their words, and posts written for each channel you ticked.
Run it again next month.
The same business measures the same topics, so this month's plan compares with last month's.
Business details stay private, and off shared links unless you switch them on.
Who it reaches. What to say.When to move.
Every marketing plan, and every idea run that asks for one, comes with a brief written from what the run measured. It moves on measured gates, not dates.
Campaign brief
built on the angle that landed · re-run to check the gatesThe plan
Position the campaign on “Class divide and exploitation of trust in domestic workers”, and lead the materials with the trust breach, framed in the audience's own domestic terms. Seed the communities first; re-run monthly to see if the next step is due.
What to do — the campaign brief
“Class divide and exploitation of trust in domestic workers” is the angle this run measured strongest, and the classifier read what lands on it and what grates.
The conversation is already running in 2 named communities this run read directly.
Whatever year you are ready, aim at the months this subject runs hot: across 3 years of the “domestic thriller” search series, November, December and January run above the yearly average.
Plus Audience definition (media buying) and Creator partnerships (digital & partnerships), each with its own named targets.
When to escalate — gates, not dates
join the named rooms in their own language
Gate: The seeded rooms grow the lead angle's share of kept conversation across a re-run.
put material where the lead angle concentrates
Gate: The lead angle's momentum reads rising without the conversation souring.
commit the campaign against your release frame
Gate: A stated release month clears the crowded stretch on the release calendar.
Test an idea.Get an answer you can check.
A film, video, post or podcast you are deciding whether to make. Describe it and name the comps; every claim that comes back links to its source.
Money moves first. Proof shows up last.
Budgets and production time are committed before there’s any evidence an audience exists. Celestia brings that evidence in while it can still change the decision.
New run
Title
Logline
The idea, the script, the comps.
A title, a logline and one comp are enough to start. Scripts are never used to train a model and are deleted after 24 hours.
Angles
4 measured, from script brief- Class divide and exploitation of trust in domestic workerspositive · ▲ rising · 432 items
- Gaslighting and psychological manipulation inside the homeupset by the subject · → stable · 37 items
- Female-led crime and domestic thrillersmixed · ▲ rising · 282 items
- Invisible labor and domestic-worker visibilityunclear · — unknown · 1 item
Class divide and exploitation of trust in domestic workers
theme · read from script briefpositive · ▲ rising · 432 items
People read this as a thriller premise they already live inside. The recurring argument is that the power imbalance needs no explaining — viewers supply it from experience — so the tension lands immediately and the scrutiny falls on whether a story earns its ending rather than on whether the setup is plausible.
What lands
- The employer/worker imbalance reads as immediately legible, no setup required
- Domestic settings turning threatening without anything supernatural
- Protagonists whose competence is disbelieved by everyone above them
What puts them off
- Endings that resolve the class conflict too neatly
- Suspicion of stories that treat servitude as texture rather than subject
What people say
- Viewers frame the tension as 'invisible abuse — shift work as a weapon'c1
Coverage & sources
how representative this read is — what was read, and what was counted- Topics analyzed:
- 4
- Returned signal:
- 4 / 4
- Tool calls:
- 43
4 sources
11 Wikipedia pageviews · 2 TMDB lookups · 1 Box Office Mojo
Weighed the mood of 3 angles. Too few dated posts on too few platforms to vote on direction.
What each grade was read from
752on-topic items1,350likes and upvotes weighing the mood1search curves1page-view histories
3search and page-view trajectories
1page-view histories
2lookups2discussion reads
By platform
Sentiment, momentum and attention are Celestia's own analysis and are not from YouTube. YouTube comments are read during a run; their text, likes and view counts are not kept.
One engine, two tracks
- Formats
- Film / TVFeature film · Limited series · Ongoing series · Premium podcast · Premium creator content
- Creator / BusinessVideo · Short form · Blog post · Newsletter · Podcast episode
- Enough to start
- Film / TVAdd a title, a logline, and a script, comp, or synopsis.
- Creator / BusinessAdd a title, a hook, and at least one theme or reference.
- Marketing plan
- Film / TVOne section per team. Seed → Widen → Open.
- Creator / BusinessWhere to post, who to pitch. Post → Read → Follow up.
Change your mind.Keep the evidence.
Try a different budget, release window or source mix, and every grade recomputes from the evidence already gathered. The original is never touched.
See how much rested on what.
Here the loudest source is hidden, and the audience grade moves with it.
Adjust a finished report
every grade recomputes instantly from evidence the run already gatheredThe questions are the same, whoever's asking.
The same engine reads for anyone whose next release, or next post, is a bet on an audience.
Businesses & brands
A marketing plan from your profile: which subjects reach your customers, and the posts to make.
Creators & creator studios
What to make next, and which angle your audience already wants.
Producers & distributors
Where the demand for a title sits, and when to release into it.
Studios & funds
Greenlight calls with the comp evidence checked, not assumed.
Three decisions. One read.
What to make, where to launch it, who to reach. Each answer changes the other two.
- Greenlight01
What should we make?
Whether the audience for this idea already exists, and how warm it is.
- Distribution02
Where and when should it launch?
Which markets and moments the demand is actually concentrated in.
- Marketing03
Who will care, and how do we reach them?
Which audiences to spend against, and which angle they respond to.
See what the audience already knows.
Bring your business or a project you’re weighing right now and we’ll run it live.