jimbojw / gendict

gendict

This project generates videos for dictionary terms. It does this through a multi-step process. For any given term, the process goes something like this:

• extract definitions for term from a wiktionary data dump;
• create a sequence of slides for the term based on the definitions;
• for each slide:
• generate an image containing the display text of the slide,
• generate an audio file containing a text-to-speech rendering of the script text of the slide,
• generate a video file consisting of the still image and audio files mentioned above.
• concatenate the video sequence into one final video.

Installation and Setup

1. Install ruby
2. Install imagemagick:

• brew install imagemagick

3. Install ffmpeg:

• brew install ffmpeg

4. Install gems:

• gem install nokogiri youtube_it
• C_INCLUDE_PATH=/usr/local/Cellar/imagemagick/6.7.7-6/include/ImageMagick gem install rmagick

5. gem install bundler && bundler install

Additionally you'll need to install any fonts in the assets/ directory of this project.

• In Finder, navigate to assets/
• For each .ttf or .otf file in this directory:
• Double-click the file to open it in Font Book
• Click the Install Font button to install it.

Grab the latest word list and definitions

gendict uses Wiktionary definitions of terms to generate the videos. Use these commands to download the latest English Wiktionary definitons.

cd dumps
curl -O http://toolserver.org/~enwikt/definitions/enwikt-defs-latest-en.tsv.gz
curl -O http://dumps.wikimedia.org/enwiktionary/latest/enwiktionary-latest-all-titles-in-ns0.gz
gunzip *.gz

Configure ENV variables

gendict relies on three environment variables to configure YouTube uploading:

• YOUTUBE_UN - your YouTube user's user name,
• YOUTUBE_PW - your YouTube user's password,
• YOUTUBE_DK - your YouTube user's developer key.

You can look up your YouTube developer key in the YouTube API Dashboard.

Generating slides, images, audio and videos

To generate the slides for a term, use the make-slides.rb script:

ruby scripts/make-slides.rb <term>

This will dump a YAML encoded sequence of slides to standard output. To see the slides in a more human-readable form, pipe this YAML data into the print-slides.rb script:

ruby scripts/make-slides.rb <term> | ruby scripts/print-slides.rb

To generate the images for the slides for a term, use the make-images.rb:

ruby scripts/make-slides.rb <term> | ruby scripts/make-images.rb

The make-images.rb script augments the incoming YAML data, adding a new image field to each slide. It pushes these updated slides as a YAML encoded sequence to standard output. The images the script generates will be placed in the terms/<term> directory.

Similarly, to generate just the audio for a set of slides, use the make-audio.rb script.

ruby scripts/make-slides.rb <term> | ruby scripts/make-audio.rb

Since the image and audio generation scripts both read and write the YAML representation of slides, they can be piped in sequence:

ruby scripts/make-slides.rb <term> | ruby scripts/make-images.rb | ruby scripts/make-audio.rb

The last script in the chain generates all the intermediate videos and then the final video. The make-video.rb script assumes that all the images and audio have already been generated. The script expects that the incoming YAML slides all have 'image' and 'audio' fields pointing to those files.

The scripts can all be chained together:

ruby scripts/make-slides.rb <term> | ruby scripts/make-images.rb | ruby scripts/make-audio.rb | ruby scripts/make-video.rb

Or they can be separated out by piping the YAML to a separate file and feeding it back in later.

ruby scripts/make-slides.rb <term> | ruby scripts/make-images.rb | ruby scripts/make-audio.rb > terms/<term>.yaml
cat terms/<term>.yaml | ruby scripts/make-video.rb

The benefit of storing the intermediate YAML is that you can re-run any of the steps without re-running the whole chain.