Markov Comic Generator

A program for generating comic strip dialog using Markov chains

Inspiration

This program was inspired by but is not otherwise connected with Josh Millard's "Calvin and Markov"

Why?

For fun.

Noteworthy files and folders in this repository

• generator.py: The Python module responsible for generating Markov chains
• main.py: The Python script responsible for everything else
• LICENSE: The copyright license governing the code (fonts and images are under separate licenses)
• README.md: This file
• data: A git subtree linked to the Mimi and Eunice Transcripts git repository.
• data/fonts: The directory in which desired font(s) should be installed. Currently is linked (via git subtree) to the Nina Improved git repository.
• data/fonts/LICENSE AND ATTRIBUTION: License and attribution information for Nina's font.
• data/images: The directory in which input images should go.
• data/transcripts: The directory in which comic transcripts should go.
• data/word-bubbles: The directory in which definition files should go.
• data/sources.tsv: An optional text file containing the locations at which original, unedited comics can be found.

The Input Directory

The input directory is the directory in which all input files must be located. It defaults to the data folder but can be specified on the command line with the --indir option.

Input file formats

Common

Comments - that is, notes in the input files that are intended to be ignored by the program - are preceded by "}}". The }} and everything coming after until the next line break will be ignored.

Blank lines - either lines that are actually blank or that become blank after removal of comments - will be ignored.

ID numbers are integers used to uniquely identify the original comic strip from which images and dialog are taken. Dates of the form YYYYMMDD work well for this and are recommended.

Excluding image files, the first non-blank line of an input file is a number identifying which comic strip this file corresponds to. ID numbers in these files are only used by the program to quickly check whether the input file is in the correct format: the ID must be an integer and must be the same as the ID in the file name. It's like when Van Halen put a "no brown M&Ms" requirement in their contracts: the colors of the M&Ms were not actually important; brown M&Ms would serve as a sign that something else might be wrong.

Each character has a label. For the sake of flexibility, these labels are not hard-coded into the program. Differences in spelling will be interpreted by the program as different characters. For this reason, it is recommended that only the character's initials be used, e.g. M = Mimi and E = Eunice. Additionally, the following labels are recommended:

• MISC = miscellaneous/offscreen character dialogue (for when the character is not clearly identifiable)
• SFX = non-spoken sound effects such as BOOM or CRASH.

Labels are not case-sensitive.

All input files are either images (any common format is acceptable; PNG format is recommended) or text files. It is recommended that text files be encoded in UTF-8 and have Windows-style CR+LF newlines.

Sources

The sources.tsv file, if it exists, must be located in the root of the input directory. Each line of the file must consist of one ID number, then a tab, then some text (preferably a URL) identifying where the original comic pertaining to that particular ID can be found.

Dialog transcript files

These transcript files should be in the transcripts/ folder inside the input directory. A transcript file's name should be its ID number followed by the ".txt" file extension.

All the lines after the ID represent dialog or sound effects. These lines take the form of a character label, a colon, a space, and then whatever dialog the character speaks.

Different panels or balloons go on different lines. This just makes for a cleaner representation of the original comics.

In the case of shared dialog (e.g. two or more characters say the same thing at the same time), a separate copy of the dialog should be given to each character on two consecutive lines.

Emphasis can be added in the following ways:

• *asterisks* surrounding a word represent bold text
• /forward slashes/ surrounding a word represent italic text
• _underscores_ surrounding a word represent underlined text.

Word capitalization should be as it is in the original comic strip.

If in the original comic, a word is hyphenated in order to split it between two lines of text, it is recommended that the word be kept on a single line in the transcript and the hyphen be replaced with a soft hyphen (Unicode code point U+00AD). Soft hyphens are not normally visible in most text editors; they are used to mark a place where a line break, possibly with a visible hyphen, can be automatically inserted if one is needed.

Example

}}This first line will be blank after the comment gets removed, so it is ignored.
20101210 }}Date serves as ID number
E: WOULD YOU RATHER BE *RIGHT,* OR *HAP­PY?* }}Eu­nice i­s talk­ing. Al­so, th­e la­st wo­rd ha­s a­ so­ft hy­phen i­n th­e mid­dle. A­s do­es ev­ery wo­rd i­n thi­s com­ment. Fo­r de­monstra­tion o­f wha­t so­ft hy­phens d­o, tr­y resiz­ing thi­s wind­ow; se­e h­ow thi­s com­ment ge­ts wrap­ped i­n compar­ison t­o th­e oth­ers whi­ch do­n't ha­ve so­ft hy­phens i­n ever­y wo­rd.
M: *NEITHER!*
}}E: *NEITHER!* }}If Mimi and Eunice both said this at the same time, we would uncomment this line.
M: I'D RATHER BE *RICH!* }}Note: For some reason Mimi and Eunice speak in all caps. See for yourself: http://mimiandeunice.com/2010/12/10/right-or-happy-iii/

Image files

Each image file should be in the images/ folder inside the input directory. Its name should be its ID number followed by an extension appropriate for the format. The image should have all dialog blanked out so that empty word balloons are left.

Word bubble files

Each word bubble file should be in the word-bubbles/ folder inside the input directory. Its name should consist of its ID number followed by ".tsv".

The first non-blank line of each file should be a tab-separated list of labels of all the characters who speak.

All following lines should take the form of one character label, a tab, the X coordinate of the top left corner of the word balloon, a tab, the Y coordinate of the top left corner, a tab, the X coordinate of the bottom right corner, a tab, and the Y coordinate of the bottom right corner. As this format may change depending on the needs of the program, to maintain compatibility between program versions, any data following this will be ignored. All coordinates are integers representing pixel coordinates in the image, with (0,0) being the top left corner.

If multiple characters use the same speech bubble (saying the same thing at the same time), create separate lines in the file - one for each speaker, one line coming immediately after the other - with identical speech bubble coordinates.

Example

20101210}}The image is stored as images/20101210.png; this definition file is stored as word-bubbles/20101210.tsv
E	M }}List of speakers
}}Speaker	X	Y	X	Y
E	37	6	180	69
M	280	48	370	76
}}E	280	48	370	76 }}If this comic had Eunice sharing a speech bubble with Mimi, we would uncomment this line.
M	488	30	586	82	ABC }}The ABC will be ignored by this program version because it comes after the final coordinate. Later versions of this program might use it.

Fonts

The program will use the first usable font it finds, searching in the following order:

1. A font file specified on the command line ("--font" on the command line)
2. All font files in the "fonts" subdirectory of the input directory.
3. Nina Improved
4. Nina
5. Humor Sans
6. Tomson Talks
7. Nibby
8. Vipond Comic LC
9. Vipond Comic UC
10. Comic Neue
11. Comic Relief
12. Dekko
13. Ruji's Handwriting Font
14. Open Comic Font
15. Comic Sans MS
16. Ubuntu Titling
17. Any other font

It is planned that in the future, should a particular font not have required glyphs, a suitable replacement font will be used for only those glyphs. This is not yet implemented.