Extract a PDF Schedule or Timetable Table to JSON

A day's programme with a day banner

Note 'Tuesday 2 June' becomes a near-empty row, and the long keynote title wraps to a second row with an empty time.

PDF:
Time         Session                         Room
Tuesday 2 June
09:00–10:00  Opening keynote: the road
             ahead for the platform          Main Hall
10:15–11:00  Workshop A                       Room 2

Downloaded JSON:
[
  { "Time": "Tuesday 2 June", "Session": "",                              "Room": "" },
  { "Time": "09:00–10:00",   "Session": "Opening keynote: the road",     "Room": "" },
  { "Time": "",              "Session": "ahead for the platform",        "Room": "Main Hall" },
  { "Time": "10:15–11:00",   "Session": "Workshop A",                    "Room": "Room 2" }
]

Carry the date down and merge wrapped titles

Walk the rows once: remember the current day banner, attach it to following events, and fold continuation rows (empty time) into the event above.

const rows = JSON.parse(json);
const events = [];
let date = null;
for (const r of rows) {
  if (/\d/.test(r.Time) && !/^\d{1,2}[:.]/.test(r.Time) === false) {} // (illustrative)
  if (/^[A-Za-z]+ \d/.test(r.Time) && !r.Session) { date = r.Time; continue; } // day banner
  if (!r.Time && events.length) {                       // wrapped continuation
    events[events.length - 1].Session += " " + r.Session;
    if (r.Room) events[events.length - 1].Room = r.Room;
    continue;
  }
  events.push({ date, ...r });
}

Normalise a printed time range to ISO 8601

Combine the carried date with the printed time range and produce real start/end datetimes — explicit timezone, no guessing.

import dayjs from "dayjs";

function toISO(date, time) {           // date "Tuesday 2 June", time "09:00–10:00"
  const [start, end] = time.split(/[–-]/).map(s => s.trim());
  const d = dayjs(date.replace(/^[A-Za-z]+ /, "") + " 2026", "D MMMM YYYY");
  const mk = t => d.hour(+t.split(":")[0]).minute(+t.split(":")[1]).second(0);
  return { start: mk(start).toISOString(), end: mk(end).toISOString() };
}
toISO("Tuesday 2 June", "09:00–10:00");
// { start: "2026-06-02T09:00:00.000Z", end: "2026-06-02T10:00:00.000Z" }

Generate an .ics calendar file

Calendars import iCal, not JSON. Build .ics from the normalised events so attendees can subscribe.

import ical from "ical-generator";

const cal = ical({ name: "Conference 2026" });
for (const e of normalisedEvents) {
  cal.createEvent({
    start: new Date(e.start),
    end:   new Date(e.end),
    summary: e.Session,
    location: e.Room,
  });
}
fs.writeFileSync("programme.ics", cal.toString());

Reshape a grid timetable into per-cell events

A class timetable with days across the top flattens by row. Pivot it into one event per (time-slot, day) cell.

// rows keyed by { Time, Mon, Tue, Wed, Thu, Fri }
const days = ["Mon","Tue","Wed","Thu","Fri"];
const slots = JSON.parse(json).filter(r => r.Time && r.Time !== "Time");
const events = slots.flatMap(r =>
  days.filter(d => r[d]).map(d => ({ day: d, time: r.Time, subject: r[d] }))
);

Every value is a string, including "9:00 AM" and "2 June". The tool never parses temporal values. Convert to ISO 8601 in your code with a date library so timezone and AM/PM handling is explicit and testable.

A 'Tuesday 2 June' header that spans the row width arrives as a row with mostly empty cells. Capture it as the current date and attach it to following events, then drop the banner row (see the cookbook walk-through).

Long titles wrap to a second visual line, which becomes a separate row with an empty time cell. Merge continuation rows into the event above — an empty time column is the tell-tale that a row is a continuation.

A timetable with days across the top and time slots down the side extracts one object per time-slot row (with a key per day), not one event per cell. Pivot it in your code to get individual events (the cookbook shows the flatMap).

From page 2 onward the reprinted column header is emitted as a data row. Filter r.Time !== 'Time' (or your first column) before building events.

A scanned booklet has no text layer, so extraction returns nothing. Run PDF OCR first to add a text layer, then extract — and double-check times, since OCR can confuse 0/O and 1/l.

If the time and the session sit very close horizontally, position-based grouping can merge them into one cell. Inspect the preview; split a "09:00 Opening keynote" cell with a leading-time regex in post-processing.

A thick multi-day conference book can exceed 50 pages. Upgrade to Pro (500 pages) or extract just the schedule pages with PDF Extract Pages before running the tool.

Calendars import iCal (.ics), not JSON. The tool's job is to get the events into structured data; convert that to .ics (e.g. with ical-generator) before importing into Google Calendar, Outlook, or Apple Calendar.