Knowing the material, applying it to the work, and catching what slipped through. Most tools gesture at the first and skip the other two. Here is how the three fit together at Chitram.AI, and where to read about each.
What does it actually know, and from where?
A curated corpus built with research partners rather than scraped — primary texts, iconographic conventions, architectural proportion systems and dramaturgical frameworks, each retrievable with its attribution intact.
See what is in the corpusHow does that reach the thing I am making?
The generation module that reads the corpus at request time, so a costume brief, a temple elevation or a scene beat resolves against documented convention instead of the model’s general impression of "Indian".
Open the moduleWhat stops a mistake reaching the audience?
A review stage that flags anachronisms and departures from convention, cites what it drew on, and says plainly when the material is contested enough to need a human rather than a verdict.
How review worksThe corpus is available to the modules you already work in, so grounding is a property of the work rather than a specialist step somebody has to remember.
Rasa applied consistently across a scene; register that holds from line to line rather than drifting.
Proportion systems and regional distinctions that make a temple read as a specific tradition, not a generic one.
Attribute sets and conventions applied to design work, and checked again before anything ships.
That this settles questions of interpretation. Traditions disagree with each other, regional practice varies, and living communities have a stake that no corpus represents on their behalf. Where the material is contested, the right output is a flag and a recommendation to consult — not a confident answer.
What we do claim is narrower and checkable: that a depiction can be traced to the material behind it, and that the obvious errors get caught while they are still cheap to fix.
Bring us a brief you are unsure about. That is a more useful conversation than a demo, for both of us.