Frequently asked questions

You upload a document and describe the result you need. ELLA can audit a file without changing it, or create an accessible Word file, braille, large print, spoken audio, a DAISY talking book, or descriptions for figures. Direct PDF uploads are audit only for now: ELLA reports problems but does not convert the PDF to PDF/UA-2. You describe the outcome and ELLA chooses the route.

ELLA asks when a request could mean more than one thing. If you say "make this accessible", you might mean an audit, a Word file, braille or audio. Naming the output avoids spending pages on the wrong job.

PDF, Word (.docx), PowerPoint, Excel, the OpenDocument equivalents, RTF, EPUB, plain text, Markdown, CSV, and images in PNG, JPEG or HEIC. You can also upload existing BRF and eBraille files if you want them checked. A ZIP is accepted when a document arrives with its assets.

Upload the original when you have it. PDFs often lose or obscure structure from Word or PowerPoint, so the source file usually produces better results. Fixing the source also improves the next version of the document.

Free accounts include 10 pages a month, Pro includes 100, and Enterprise includes 1000.

It depends on the length of the document and what you asked for. An audit is simpler than turning a long book into braille or audio. ELLA shows progress while it works, and you can leave the page and come back.

ELLA will tell you it is a scan rather than quietly guessing. A scanned page has no text layer, so there is nothing for a screen reader to read and nothing to tag. Text can be recovered, but ELLA will not write that recovered text back into the file and then call the document accessible, because it cannot verify that the transcription is correct. It will tell you plainly that there is no readable text for a screen reader to use.

No, and you should be wary of any tool that says otherwise. ELLA runs real validators and will tell you what passed and what did not, but a person who knows the subject still needs to read the result before it goes to a student. This is most true for maths, complex tables, and any figure whose meaning is carried entirely by the picture.

Read the report ELLA gives you, then look at the things a validator cannot judge: does the reading order make sense read aloud, do the figure descriptions say what the figure is actually for, do complex tables still make sense linearly, and is the subject matter right. Keep your original file until you are happy with the result.

Use ELLA to prepare the material, but keep a qualified human check before anything is issued. It can find and fix a lot of mechanical problems. It cannot judge whether a description, a table or an equation gives that student the right information. The point is to spend your reviewer's time on judgement instead of on tagging tables by hand.

Yes, when the equations exist as real maths in the source. It converts them into maths a screen reader can actually speak. Equations that only exist as pictures are a harder problem, and ELLA will say so rather than describing an image and calling it an equation.

Unified English Braille, the standard across the UK, Ireland, the United States, Canada, Australia and New Zealand. Three systems are supported: UEB grade 1 (uncontracted), UEB grade 2 (contracted), and UEB technical braille for maths and science. ELLA states which one it used on every braille output and flags anything that needed a judgement call. Nemeth is not supported, so if your reader needs it, ELLA is not the right tool for that job.

eBraille is a newer packaged format for braille publications, with navigation and proper metadata. It requires the title, the author, the language, the copyright year, and which braille system you want, and the file is not valid without them. ELLA asks because it will not invent a copyright year for someone else's book.

Ask ELLA that question directly, without uploading anything. It will talk through how to triage a pile of material: which formats cause students the most trouble, what gives the biggest improvement for the least work, and what is worth fixing at source instead of remediating every year.

No. Your uploads are processed to do the work you asked for and are not used as training data.

Email support@autodots.ai. If a specific document produced a bad result, say which file and what you asked for, because that is usually enough to reproduce it.