What files does the checker accept?
Braille2000 .B2K and .BML files. A .B2K carries both the rendered braille text and the underlying markup, which is what lets the checker reason about structure — tables, boxes, page numbering — rather than only reading cells. You can add the tactile graphics alongside it too: a CorelDRAW .cdr, or a PDF export of one.
Why is a PDF export better than the .cdr?
Because of what each one can give up. A .cdr keeps its page contents in CorelDRAW’s own binary format, which it would take CorelDRAW itself to read — so the page count is genuinely all the checker can take from it. A PDF exported from the same artwork carries live braille text, so every graphic page can be checked on its own: the braille on it and its page numbering. If you have both, add the PDF.
What if my .cdr cannot be read?
Files saved by CorelDRAW X4 or later are read directly. An older file uses a different internal format, and rather than guess at a page count the checker reports that it could not read it. Export a PDF of the artwork instead — that works whatever version drew it, and it is the better input anyway.
Does this replace proofreading?
No. It catches mechanical and formatting errors — an emphasis passage that was never terminated, a Special Symbols page out of braille order, a table cell crossing a column boundary. It cannot tell you whether a heading is really a heading or whether a graphic was described well. A proofreader still reads the book; the checker makes sure they are not spending that time on things a machine can find.
Do you keep my files?
No. Uploads are written to a temporary directory, checked, and deleted before the response is sent back to you. Nothing is retained and nothing is logged except the number of findings.
How were the rules decided?
They were measured against 297 real transcribed volumes rather than read off a rulebook in the abstract. Which symbols must appear on a Special Symbols page, for example, was derived from how often each one actually appears across the 175 files that have such a page. Rules that produced noise on known-good files were tightened or dropped.
Will it flag things that are actually correct?
Less than you would expect, because the rules were tuned against real files until known-good volumes came back clean — 189 of the 297 test volumes report nothing at all. Where a rule is less certain it reports a warning rather than an error. If you do hit a false positive, tell us which rule and we will look at it against the corpus.
How much does it cost?
It is free while in early access. It will become a paid feature later, and we would rather say so now than surprise you. You need a free QBT account to use it, which also gives you 25 free pages in the braille converter.
How fast is it?
A hundred-page volume is checked in well under a second. You are not waiting on it.