Work | About | Blog | Flow | Manifesto | Code
Tags | Index
 
Trace: » Faber Finds generative book covers


Faber Finds generative book covers

Print on demand (POD) is becoming more mainstream by the day and new(ish) contenders are popping up everywhere, slowly covering more and more parts of the market. Recently the renowned English publisher Faber & Faber has also joined that field. However, unlike the other players who are largely focused on the self-publishing game, Faber's plans were to bring POD to that part of the market, which due to the traditional ways of the publishing business has ceased to be a market at all: out-of-print books. Faber Finds is the name of said service and has just launched with an initial repertoire of 100+ titles, with more being added all the time.A year ago Faber & Faber commissioned me to help with the design of a software system to generate complete & print ready book covers for their new imprint. The challenge proved to be more of a creative than a technical one, as the task given was to build a “design machine” which would be flexible enough to generate a very large (theoretically infinite) number of unique designs, one for each single book ever printed in this range, within the agreed boundaries set by the art direction(s) of Faber design team. The imprint has currently 4 genres, each with its own slightly varied style and rules I needed to take care of.

Faber Finds collection

Faber Finds collection

Faber Finds collection

Some of the first printed books with the fantastic custom made B-HMMND font
by Michael C. Place @ Wearebuild.

Border designs

Faber also commissioned the super talented Canadian typographer Marian Bantjes to create four designs used as templates for the desired look & feel of the borders styles of each of the different genres offered by the imprint. Each of her design routes then needed to be abstracted, decomposed/split into smaller elements & shapes, parametrized and generally reverse engineered conceptually. The shapes would then become micro templates, or rather, form a shape vocabulary for the complex borders. Only once I understood all the rules and nature of the design elements used on all levels, I could start building a generative solution which would introduce variations at certain points of the design process, manage and judge them automatically.This initial part of the process included things like identifying the 5 levels of symmetry used within Marian's sketches, experimenting with minimum and maximum border widths, exploring individual symmetry options & limitations for each of the shape elements below, finding the right density range of shapes used per border quadrant, ensuring author names and titles are correctly word wrapped whilst not obscured by the borders, auto-adjusting the font size for text on the spine based on the number of pages vs. length of text etc. In total we isolated over 35 such rules and parameters…

Extracted & re-generated shape vocabulary

Extracted & re-generated basic shape vocabulary from Marian Bantjes' sketches.

Early sketch with debug markers

Early sketch with markers showing registration and symmetry points for each shape element.
(Btw. Umberto Eco is NOT published by Faber Finds)

Doing it in software

Generative design systems often work on the premise of extrapolating a given design idea/art direction. Because our aim was to create such a system (rather than a single one-off design), it was important to find the extreme cases and boundaries of expressions possible and shape them. This idea is going back to the multi-dimensional design space of possibilities. Finding appropriate values to these design parameters required a phase of constant experimentation and conversations with Faber's design team - these collaboratively agreed boundary values then became the encoded art direction within the software.Generating the borders was just one, if major, task of the final solution, though. The custom software written in Processing, straight Java and PHP works as an internal webservice at Faber which receives new batch orders and then generates complete, print ready PDF files with all copy, branding, spine, ISBN, barcode and optional high-res JPG preview using the book details supplied. Generating a single cover only takes about 1 second, but due to its iterative and semi-random nature can sometime require hundreds of attempts until a “valid” design is created which is judged to be “on brand” by software itself.

A fully generated book jacket with spine, barcode, back copy and ISBN

A fully generated book jacket with spine, barcode, back copy and ISBN

Under the bonnet

In addition to Processing's built-in PDF generation functionality it was a great help to work directly with the underlying iText PDF library, e.g. in order to generate the EAN13 barcode and add meta data to the saved PDF files. For both things I needed to patch PGraphicsPDF library to grant me lower level access to iText.There're also some useful unit conversion functions for dealing with page sizes in PDF in the toxiclibs SVN… In another post we will be taking a more proper look under the bonnet of the whole generator application…

Faber Finds collection

For now: Happy reading!

, , , , , , , , , ,

Discussion

Joe Clark, 2008/07/29 18:25:
Do you think you could upgrade the software to output tagged PDFs? I’m assuming it doesn’t do so already and that it is producing live text, not images, which is probably an erroneous assumption. Tagged PDFs would make electronic versions of these custom books at least passably adaptable for print-disabled users, as they are often called.
Karsten Schmidt, 2008/07/29 18:27:
Hi Joe, aren't you talking about producing the actual books as tagged PDF? AFAIK none of these books are available as e-books, but only as hardcopy and so whilst probably a good idea, it doesn't really apply here (unfortunately!). Furthermore, my software only generates the PDF for the book jacket, not the actual book pages... Does this clarify things?
Jane Gorton, 2008/07/31 01:40:
I recently read about Expresso Books, a book printing system by ondemandbooks. Is this the same system or does your book printer focus only on out of print books?
Karsten Schmidt, 2008/08/09 15:31:
The software I developed is fairly independent on the actual printing system since the only output are print ready PDF files. Though, Faber Finds is purely focusing on out-of-print books and not the usual "vanity press" type offering.
Carlos Meirinho Carrilho Rito, 2008/08/07 01:59:
www.paulinas.pt
Jens Alfke, 2008/08/09 09:05:
This is neat, but I'm a bit unsure what you mean by "one [design] for each single book ever printed". Do you mean one for every _title_, or one for every _physical printed copy_? I at first assumed the latter, which would be an amazing idea -- every individual copy unique, like a snowflake or a Cabbage Patch doll -- but on closer reading I think you mean that every copy printed of a given title has the same design. (None of the photos show more than one copy of any title.) This is still a very clever way to differentiate the covers without having to invest in design for each title, but it doesn't take advantage of the unique possibilities of print-on-demand. Would it be possible to make each copy different, or does the current POD infrastructure not allow for this?
Karsten Schmidt, 2008/08/09 15:28:
I actually did mean the latter too, every physical printed copy unique, leaving the era of mass production behind - the software was built for this exact context. Though having said this, I really can't tell if Faber are following fully through with this plan. In the early stages of the project we also talked about a special dedication message on the first page of the book ("Printed for YOUR NAME etc."), but there were production issues with that (not sure why, but am guessing the book covers have a slightly different print process than the book contents). Hope that helps!
website designs, 2008/08/26 21:10:
Fantastically ugly.
Anon, 2008/08/26 22:54:
You people that keep saying they're ugly need to get your eyes checked asap.
David, 2008/08/31 19:57:
A wonderful initiative by Faber & Faber. My comment is not on design but on punctuation, which I hope you agree is a justifiable topic for a publisher's website, and though dismissed by some as pedantic is as critical as is attention to design detail. Your intro above says, "However, unlike the other players who are largely focused on the self-publishing game, Faber's plans were to bring POD to that part of the market, which due to the traditional ways of the publishing business has ceased to be a market at all: out-of-print books." I applaud your intent to avoid the commonplace Britishism of using 'which' for a defining clause when 'that' should be employed, but I fear you maybe haven't got it quite right here. Certainly the sentence reads strangely, and perhaps is due to your wanting to avoid inelegantly using 'that' too many times. You would surely have ready access to a competent copy editor who could check this, but I think it should be punctuated thus: "... Faber's plans were to bring POD to that part of the market which, due to the traditional ways of the publishing business, has ceased to be a market at all: out-of-print books. (And perhaps better to use a semicolon at the end?)
Leia, 2008/09/04 15:15:
They're difficult to read because the typeface and borders are too decorative and distract from one another so the overall result is a complete mess and your eye doesn't know where to look. This design doesn't work for a book cover because consumers will find it takes too long to work out what they're looking at and move on without bothering to even pick up the book. The idea is a good one in terms of putting covers cheaply on POD titles but the design fails these books.
Enter your comment:
Last modified: 2008/07/21 05:01