DNAML releases PDF to ePub

[mtravellerh]

This seems to be a very complicated PDF, so it may be a special case problem. Nonetheless, I would like to try out the software for myself!

[Grauheim]

No, truly, this program is absolutely awful. I tried converting three more PDF's to experiment. The first two failed to covert with Lua errors reported in PDFtoEPUB's log file. The last did convert, but had even worse formatting problems than the one I posted. Bullet lists weren't formatted correctly, and at times, the formatting randomly switched to centered mode instead of justified.

I think that the program can handle bitmapped images (e.g., photos or screenshots), but not vector images (e.g., engineering diagrams or graphs.) It simply removes the latter.

I suspect that many (most?) people who are interested in this kind of program would be using it for text books or papers, and so will be very unhappy with it.

Slightly refining what I said in my earlier post, it probably works ok for PDFs involving just simple text and photographs. But anything involving graphs, diagrams or mathematics it cannot handle.

[frabjous]

Thanks for letting us know!

If I were you, I'd show them your results and demand your money back.

[PieOPah]

Would you be able to do a test against a PDF composed mainly of text?

I have a couple of PDFs that I would like to convert so would be interested to know how it handles them...

[kennyc]

WRT the initial posting:

I don't believe it.

Of course I'm an engineer with scientist leanings --- skeptical by nature you might say.

[Slite]

And also, can this piece of software stop BEFORE it converts to ePub? Or does it also save the file it creates from the pdf that it then converts to ePub?

Why I ask is if it would be possible to use this to convert a pdf to ePub, throw the ePub away, take the HTML file and edit that, and then use frinstance Sigil instead to create the epub?

[igorsk]

Grauheim, can you check how PDF Transformer works with your files? (trial version available)

[frabjous]

Originally Posted by Slite

And also, can this piece of software stop BEFORE it converts to ePub? Or does it also save the file it creates from the pdf that it then converts to ePub?

Why I ask is if it would be possible to use this to convert a pdf to ePub, throw the ePub away, take the HTML file and edit that, and then use frinstance Sigil instead to create the epub?

You can tell from the screenshots on their site that the software has both HTML and ePub output options. So you could use it to convert PDF to HTML, and then use Sigil, or whatever else, to create the ePub. (Or for that matter, use Sigil to edit the ePub, right?)

[Slite]

Originally Posted by frabjous

You can tell from the screenshots on their site that the software has both HTML and ePub output options. So you could use it to convert PDF to HTML, and then use Sigil, or whatever else, to create the ePub. (Or for that matter, use Sigil to edit the ePub, right?)

Ah, nice... then it MIGHT be useful after all, as long as it can handle pdf's containing mostly text decently

[Grauheim]

Slite: yes, I did try the HTML version. It's in some ways worse than the ePub, because even the bitmapped pictures are thrown away. The result is only text, with none of the figures or math preserved. Of course, if your goal is to extract the text and hand-edit it, then this will work.

igorsk: Sure! I'll give PDFConverter a try.

PieOPah: I tried converting Bleak House from the Gutenberg project (fairly plain text). PDFtoEPUB failed with internal Lua errors. I tried extracting a 5 page segment, and even that failed. I now give up.

frabjous: you better believe I've written to them for a refund!

[wgrimm]

Originally Posted by Slite

Having tried just about every PDF -> anything conversion software there is. This claim is a load of bull as far as I am concerned.

The day they DO release a piece of software that can correctly convert a pdf to whatever else format correctly, I will buy it, almost at any cost!

Well, I have tried all other existing software for pdf conversion, and the best is Gemini but nothing is perfect. And Gemini is $150. Anyways, a far easier solution is to forget time-wasting conversions by purchasing a Kindle DX.

[wgrimm]

Originally Posted by Grauheim

Well I went ahead and bought this. My recommendation: avoid at all costs. Perhaps it does fine on documents that are simple text, but complex pdf is a disaster.

Not surprised. I am not sure if it would be possible to write software that would correctly transform complex pdf docs perfectly into an epub or html format. The best pdf conversion software I have found (haven't tested this new one yet) is Gemini (by Iceni). But the far better solution is to buy an 8X10 reader that natively displays pdf- only one now is the Kindle DX, but others will be on the market soon.

Having attempted pdf conversions for years and been mostly dissatisfied, I won't waste any more time trying.

[frabjous]

When I have PDFs that are improperly sized for my reader, I've had by far the most luck with tools that either modify, or rasterize and then modify the PDF rather than converting to some other text based format: between soPDF, PDFLRF, PaperCrop, Rasterfarian and PDFread, I can usually make any PDF fit on my 505, even highly technical journal articles. But these tools work in a very different way, preserving the exact look of the PDF but cutting it up.

Heck even if my target device were a kindle, I'd seriously consider using PDFLRF and then convert the LRF to mobi afterwards.

[chenbin.sh]

Hi, I'm the lead developer of the pdftoepub.

Please upgrade to the new version (just run the installer and override the old version). The new version will run the adobe epubcheck automatically after conversion. So I assume the html syntax error should never happen.

The attached "HR GI 2009 - Camera Ready-your-rule.epub" is the epub converted by new version.

The real power of pdftoepub is it gives you a powerful script programming language to manipulate PDF texts. As a demostration, I add 3 lines in the rule.lua, and produced "HR GI 2009 - Camera Ready-your-rule.epub" to handle the Reference section better.

pdftoepub is a enterprise expert system which let the customer control everything in PDF (except vectors) with lua programming languae. Until now, I can not find this feature in mobicreator, stanza, calibre...

If anyone finds some pdf can NOT be handled properly by pdftoepub, please contact our customer support at http://www.dnaml.com/contactus.asp?product=PDF+to+ePub.

Our company are very serious about customer support, so I usually read most emails from customers.

Originally Posted by Grauheim

Well I went ahead and bought this. My recommendation: avoid at all costs. Perhaps it does fine on documents that are simple text, but complex pdf is a disaster.

I tried it on a fairly typical computer science paper in two-column ACM format. I've attached the pdf and the resulting ePub so you can look for yourselves.

The first couple of pages are fine, although it messed up the table of author names. No big deal.

It correctly detected that there are two columns and got the flow right.

The first column-spanning figure on page 2 was done correctly. So far so good.

The graphs on pages 3 and 4 are a complete mess. Rather than treating them as images, it extracted the axis labels and threw away the graph itself.

The formulae on pages 5 and onwards are thrown away.

Then about half way through the first column of page 6, the document just stops. The entire second half of the paper is unreadable. When I poke through the actual files (specifically ch0-p1.xhtml), it is apparent that there is malformed XHTML that is causing my reader to croak.

Like I said, for simple documents involving only straight text and perhaps some images, I imagine the product might work fine. For real scientific documents, however, the generated ePub is simply of no use whatsoever.

[Grauheim]

Thanks for posting, chenbin.sh. But even in the new version, the graphs and the mathematical formulae have been completely removed, and so the epub version is not useful. If you'd even converted them to images (analogously to latex2html), it would have been better than just throwing them away.