Figure Builder makes charts from your tables and diagrams from your text. It reads the table you already have (a CSV, a TSV, an XLSX), works out what each column actually is, and offers only the figure types that shape of data can honestly support: 45 types across nine families. Paste a paragraph instead and it draws one of 15 diagram types in your own words. Exported as SVG, PNG up to 1200 DPI, or true vector PDF. One credit reads a table or a text; every layout and every export after that is free.
Why the figures are the part that gets rushed
The figures are usually made last, in whatever tool was already open, the week the deadline lands. A spreadsheet chart carries its own default palette, its own grey plot background and its own font, none of which belong in a journal. Reviewers notice, and a figure that has to be redrawn during revision is a figure that was drawn twice.
The other failure is quieter and worse: a chart that is drawn correctly from data that cannot support it. Error bars taken from whatever numeric column was nearest. A trend line through categories that have no order. A pie chart of values that do not sum to a whole. The tool that drew it did not check, because it was never asked to.
What it does
Reads the table's shape first — which columns are categories, which are numbers, which are dates, how many rows there are — and then offers the figure types that fit it. Specialist figures such as ROC curves, volcano plots or CONSORT diagrams are offered only when a column is named for them. Of 45 types in the catalogue, a given table typically supports twenty-odd. The rest are not offered, rather than offered and wrong.
Every option is a live thumbnail drawn from your own table, grouped by family — comparison, distribution, relationship, trend, composition, matrix, model evaluation, clinical, and flow diagrams including PRISMA and CONSORT. You are choosing between pictures of your data, not between names.
Statistics are computed, not approximated. Quartiles, histogram bins, regression lines and confidence bands are calculated from your values using published methods, and each statistical function is tested against a worked example from the literature rather than against its own output. ORBIT draws your results; it does not run your statistical tests. A significance bracket shows the label you give it, forest plots pool by fixed effect only, and Kaplan-Meier curves carry no log-rank p-value. Those choices belong to you and your statistician.
Error bars are drawn only from a column you point at. A figure claiming a measurement nobody made is worse than no figure, so nothing is inferred from a spare numeric column.
Exports as SVG, PNG at 300, 600 or 1200 DPI, and PDF. The PDF is genuine vector output — text stays selectable text, lines stay lines — so it survives being placed, scaled and printed.
Turns text into a diagram: paste a methods paragraph, a list of themes or a project plan, and get a flowchart, sequence, cycle, mind map, timeline, tree, fishbone, pipeline, hypothesis model, comparison matrix, 2×2 matrix, Venn diagram or graphical abstract, or a PRISMA 2020 or CONSORT 2025 flow diagram drawn to the official template, or let it choose. Edit it in place for free: retype a box, add or delete boxes, draw arrows, drag them. In a PRISMA or CONSORT diagram, a count your text does not give is shown as n = ? (never worked out for you), and counts that do not add up are flagged with the sum. Labels are phrases from your text, and a number, date or title appears only if your text states it. A tree can be switched to a mind map or a flowchart for free, and any figure can be downloaded as a PowerPoint slide.
Colour safety is measured, not eyeballed. Every palette is checked for the two common kinds of colour blindness before it ships, and the builder tells you when a figure has more series than colour alone can separate. Choose the default, Okabe-Ito, or a greyscale palette that adds hatching for print. Text is set in a sans-serif face, which many journals ask for in figures.
What it costs
1 credit per table, taken when the table is analysed, and 1 credit per text turned into a diagram. Everything after that — every figure type, every re-export, every size — is free. If no figure type fits the table, nothing is charged at all, and re-running the same table inside a day returns the same analysis rather than charging for it twice.
On the free Base plan you get 5 free credits and 10 searches in each free tool every month. Starter (₹799 a month or ₹8,630 a year) and Ultra (₹1,499 a month or ₹16,190 a year) give 10 and 25 credits every 30 days, 20 and 50 searches, and a Library; they renew on their own until you turn auto-renew off, or you can buy a 30-day pass that does not renew. Top-ups cost ₹150 a credit in packs of 5, 10 or 15 and do not expire while the account is active. Most things cost one credit; a readiness check against a named journal and a conversion to your own uploaded template cost 2. Plan credits cover all of it, and no card is kept on file. The exception is a university email address: until its institution sets up ORBIT, it joins a waitlist for that institution instead of getting free credits, and it can still buy credits.
What happens to the data
The file is parsed in your browser. What crosses the network is the schema — the column names, the inferred types, the row count — and never the cell values. That is enforced by a test over the schema module, not by a sentence on this page. Figures are rendered locally from your own data, and nothing is written to disk. For a diagram, the text you paste is sent to our AI to find its structure, and it is not stored.
What it does not do
It does not offer JPEG, TIFF or EPS. JPEG is lossy and wrong for line art; TIFF and EPS are legacy submission formats that a vector PDF now satisfies at almost every publisher. If a journal insists on TIFF, open the PDF and export one.
It does not interpret your results, choose your argument, or tell you which figure to publish. It draws what the data supports and leaves the judgement where it belongs.
It does not clean your data. A table with a broken column comes back with fewer figure types offered, not with a guess at what you meant.