Upload your own material into Novus Learn

Open /upload and add a PDF, EPUB, Word document, plain text file, or Markdown file. The one distinction that decides everything is whether the file contains real text or a picture of text. A born-digital PDF is full of extractable sentences; a scan of a printed page is an image, and there is nothing in it for a text parser to lift.

If you have a scanned document, run it through OCR first and upload the result. Novus Learn will not silently invent text where it found none — a scanned upload produces an empty or near-empty project rather than a hallucinated one, which is the correct behaviour and an unhelpful surprise if you did not expect it.

The file is parsed in the browser and the resulting project is stored in versioned IndexedDB snapshots on this device. That is the whole architecture. There is no upload in the usual sense, no account holding your documents, and no server-side copy to worry about or request the deletion of.

The flip side is the same one that applies to every local-first tool: this browser is the only place your project exists. Clearing site data removes it, and another machine will not have it. Keep the original file wherever you normally keep it.

Long PDFs are where extraction quality varies most. Multi-column layouts, footnotes, headers, tables, and figure captions all interleave in ways that can scramble reading order. Skim the extracted claims for a section you know well and confirm the text arrived intact.

If a document extracts badly, it is usually the layout rather than the length. A two-column academic PDF and a single-column report of the same length behave very differently, and it is worth knowing which you have before building a fifty-card deck on top of it.

From here it is the usual seven artifacts: summary, study notes, flashcards, quiz, concept map, guided lesson, and visuals. For a textbook chapter, cards and a quiz are the natural output. For a specification or a handbook, notes and the concept map tend to be more useful, because the thing you actually need is the shape of the rules rather than recall of individual sentences.

Every artifact carries the passage it came from, so when a card looks wrong you can settle it against the document rather than arguing with your own memory of it.

For long-form course material, the /textbook surface is chapter-aware in a way a flat upload is not. If you are working through a book rather than a chapter, start there instead — the structure it preserves is exactly the structure you want the concept map and guided lesson to reflect.

The two surfaces are complementary rather than competing: upload a chapter you were handed as a PDF, use /textbook for the book you are working through end to end.

If the reason you uploaded the document was to teach from it, /teachers turns the project into a printable lesson: learning objectives, key points, a worksheet with an answer key, review cards, and source citations. It prints or saves to PDF from the browser with nothing installed and nothing uploaded.

Because the lesson is assembled from the document's own sentences, the handout says what the source says. For anyone teaching from material they are responsible for getting right, that property is worth more than a nicer template.