Kindle Voyage, released October 2014.

• No root hacks known as of now.
• Better resolution than paperwhite, can still not natively display epub files
• calibre works with the Voyage

Nowadays, you might want to consider a Kobo. It is on pair with e-ink technology, reads epub directly and much more hackable, see here for custom firmware: preining.info mobileread.com also Japanese dictionary exists

• https://standardebooks.org/
• https://www.gutenberg.org/
• https://www.aozora.gr.jp/ – Japanese texts

Reading own Japanese content in traditional vertical style (縦書き), is possible. Calibre can not generate vertical writing documents, but files generated with other tools can be managed/uploaded to the Voyage. With PDF files it is possible to read Japanese vertically, but marking Kanji and looking them up in the dictionary does not work. The following procedure produces .mobi files, appropriate for reading and marking.

• The Voyage provides by default a Japanese→Japanese dictionary
• jmdict-kindle is a free japanese→english dictionary, works nicely on Voyage and Paperwhite
• alternative dictionary: 「青Ｐ」(aop), shareware

As input, UTF8 textfiles can be used. Or the .zip files obtained from aozora/青空, a library with free Japanese content, kind of Japanese “project gutenberg”. The following method provides these benefits:

1. get the Aozora pub software and extract it: https://www18.atwiki.jp/hmdev/pages/21.html (here AozoraEpub3-1.1.0b32.zip)
2. get the kindlegen software and extract it in the same directory: https://www.amazon.com/gp/feature.html?docId=1000765211 (here using kindlegen_linux_2.6_i386_v2_9.tar.gz on Fedora19 64bit)
3. start aozora: “java -jar AozoraEpub3.jar” (here OpenJDK is used, java-1.7.0-openjdk)
4. select “.mobi” as output format
5. now select the input file and get the .mobi file generated
   1. obtain a data file from Aozora, for example https://www.aozora.gr.jp/cards/000311/files/4167_ruby_14660.zip . For these, input type MS932 should be configured.
   2. or use a UTF8 textfile with your content as input

Getting UTF8 txt files displayed on the paperwhite and beeing able to use the japanese→japanese dictionary:

• start it using “java -jar AOP3.jar” (openjdk from Fedora20 64bit works)
• configure input file, UTF8
• configure the 3 fonts, i.e. “/usr/share/fonts/sazanami/mincho/sazanami-mincho.ttf” from package sazanami-mincho-fonts. Those have to be embedded in the pdf file.
• as font sizes, configure 30, 7.5 and 10
• in tab テキスト configure custom resolution of the Voyage
• then a PDF file can be generated
  • cons: generated PDF is not nice to read on the Voyage
  • pros: but text can be marked and looked up in the dictionary

• OCRmyPDF https://github.com/ocrmypdf/OCRmyPDF – Generates a searchable PDF/A file from a regular PDF
• create epub/mobi from image files: kcc.iosphe.re
• “OCRmyPDF -l deu german_input.pdf output.pdf” creates a textlayer in output.pdf for cut and pasting
• preparing a text with fuligana, so AozoraEpub3.jar creates a fuligana .mobi file:

cat scifi.txt | kakasi -JH -i utf8 -f | \
  sed -e 's,\[,《,g' -e 's,\],》,g' >scifi_furigana.txt