config.yml

# yaml-language-server: $schema=./ConfigSchema.json

# MagiaTimeline config

# This is the configuration file that defines the parameters required for the main program to run. 
# In order for the program to work for your video, you should adjust the parameters here. 
# The file is written in YAML format and constrained by JSON Schema in ConfigSchema.json. 
# However, you do not need to be an expert of YAML or JSON Schema to modify this file. 
# Any characters after a "#" in a line are comments that will be ignored by the program. 
# There will be comments for entries below to help you understand and modify the parameters. 
# If you are still not sure how to do it, just "do as Romans do": imitate existing defaults. 
# Take it easy, the program will verify this file before running, so you are unlikely to 
# harm the program with ill-formed parameters. 

# Rectangles

# Any rectangle (or "rect") to be mentioned below defines a rectangular area in a frame. 
# It is denoted in a tuple of four decimals from 0.0 to 1.0, [left, right, top, bottom]. 
# Each decimal represents a width or height in ratio to the whole frame. 
# For example, [0.0, 1.0, 0.0, 1.0] represents the whole frame, 
# and [0.50, 1.0, 0.0, 1.0] represents the right half of the frame. 

# source: path(s) of source video files
# You may specify multiple source files for batching.
# They will be processed with the same settings.
source:
- src.mp4
# - src2.mp4

# destination: path(s) of destination subtitle files with no extension
# Include only the name of the file, without the extension,
# because they may also be used by extra jobs.
# There must be equal number of source(s) and destination(s). 
destination:
- MagiaTimelineOutput
# - MagiaTimelineOutput2

# assTemplate: template ass file defining the header of destination ass file, in asst format
# The default template should work for most cases. 
assTemplate: template.asst

# enableOpenCL: enable OpenCL acceleration
# If you have a graphics card or other accelerators, enabling this option may speed up the program.
# However, if you encounter problems or the program runs slower, you should disable this option.
enableOpenCL: false

# contentRect: defines the rectangle that wraps the content of a frame, cutting away black bars around
# If your video has black bars around the canvas, you should set this rectangle to cut them away. 
# For example, if your video has 1% of black bar on left and right and 9% on top and bottom, 
# then you should set this to [0.01, 0.99, 0.09, 0.91]. 
contentRect: [0.0, 1.0, 0.0, 1.0]

# Strategies and Presets

# MagiaTimeline supports different types of videos by providing sets of algorithms called Strategies. 
# Each Strategy comes with a set of tunable parameters and several presets of them. 
# You need to choose which Strategy and which preset of it to use. 
# For general purpose, bcs is the recommended Strategy.
# It is likely to function for a variety of videos with default parameters. 
# There are also other legacy strategies for specific types of videos. 
# They are more accurate and efficient for their specific types of videos, but may not work for others.
# They also require more tuning of parameters to work properly.

# strategy: which algorithm set to use
# - mr: MagiaRecord quest video
# - mr-s0: MagiaRecord Scene0 film
# - lcb: LimbusCompany quest video
# - lcb-mech: LimbusCompany mechanics introduction video
# - pkm: PekemonEmerald quest video
# - prk: Parako anime conte
# - bdr: BanG Dream story video
# - otl: Outlined text
# - bcs: Box colour stat (general-purpose)
strategy: bcs

# preset: which preset of parameters under a strategy to use
preset: default

# MagiaRecord quest video
mr:
  default:
    dialogOutlineRect: [0.25, 0.75, 0.60, 0.95]
    dialogBgRect: [0.3125, 0.6797, 0.7264, 0.8784]
    blackscreenRect: [0.15, 0.85, 0.00, 1.00]
    whitescreenRect: [0.15, 0.65, 0.00, 1.00]
    cgSubAboveRect: [0.0, 1.0, 0.60, 0.65]
    cgSubBorderRect: [0.0, 1.0, 0.65, 0.70]
    cgSubBelowRect: [0.0, 1.0, 0.70, 0.75]
    cgSubTextRect: [0.3, 0.7, 0.70, 1.00]
  narrow:
    dialogOutlineRect: [0.21, 0.79, 0.60, 0.97]
    dialogBgRect: [0.27, 0.73, 0.75, 0.90]
    blackscreenRect: [0.15, 0.85, 0.00, 1.00]
    whitescreenRect: [0.15, 0.65, 0.00, 1.00]
    cgSubAboveRect: [0.0, 1.0, 0.67, 0.72]
    cgSubBorderRect: [0.0, 1.0, 0.72, 0.77]
    cgSubBelowRect: [0.0, 1.0, 0.77, 0.82]
    cgSubTextRect: [0.3, 0.7, 0.77, 1.00]
  # You can create and use your own preset, as long as it conforms to the format of its strategy. 
  myownpreset: 
    dialogOutlineRect: [0.25, 0.75, 0.60, 0.95]
    dialogBgRect: [0.3125, 0.6797, 0.7264, 0.8784]
    blackscreenRect: [0.15, 0.85, 0.00, 1.00]
    whitescreenRect: [0.15, 0.65, 0.00, 1.00]
    cgSubAboveRect: [0.0, 1.0, 0.60, 0.65]
    cgSubBorderRect: [0.0, 1.0, 0.65, 0.70]
    cgSubBelowRect: [0.0, 1.0, 0.70, 0.75]
    cgSubTextRect: [0.3, 0.7, 0.70, 1.00]

# MagiaRecord Scene0 film
mr-s0:
  default:
    dialogRect: [0.35, 0.65, 0.74, 0.93]
    blackscreenRect: [0.15, 0.85, 0.00, 1.00]
  narrow:
    dialogRect: [0.32, 0.68, 0.76, 0.96]
    blackscreenRect: [0.15, 0.85, 0.00, 1.00]

# LimbusCompany quest video
lcb:
  default:
    dialogRect: [0.18, 0.82, 0.79, 0.97]

# LimbusCompany mechanics introduction video
lcb-mech:
  default:
    dialogRect: [0.30, 0.70, 0.875, 0.915]

# PekemonEmerald quest video
pkm:
  default:
    dialogBgRect: [0.05, 0.95, 0.75, 0.95]

# Parako anime conte
prk:
  default:
    dialogRect: [0.00, 1.00, 0.86, 1.00]
  vertical:
    dialogRect: [0.00, 1.00, 0.57, 0.69]

# BanG Dream story video
bdr:
  default:
    dialogRect: [0.06, 0.94, 0.82, 0.96]

# Outlined text
otl:
  yukkurirailways:
    # dialogRect: rectangle of the detected dialog area
    dialogRect: [0.00, 1.00, 0.75, 1.00]
    # fastMode: fast mode skips boundary detection for speed but may reduce accuracy
    fastMode: false
    # textWeightMin: minimum weight for areas to be considered as text
    textWeightMin: 3
    # textWeightMax: maximum weight for areas to be considered as text
    textWeightMax: 25
    # textHSVRanges: HSV ranges for areas to be considered as text
    # Each range is a list of six integers: [Hmin, Smin, Vmin, Hmax, Smax, Vmax]
    textHSVRanges:
      - [0, 0, 240, 180, 16, 255]
    # outlineWeightMax: maximum weight for areas to be considered as outlines
    outlineWeightMax: 15
    # outlineHSVRanges: HSV ranges for areas to be considered as outlines
    outlineHSVRanges:
      - [0, 0, 0, 180, 255, 16]
    # boundCompensation: extra width applied during the text-outline mutual filtering process
    boundCompensation: 4
    # sobelThreshold: threshold for the Sobel edge detection
    sobelThreshold: 250
    # nestingSuppression: suppression weight of the nested outline
    # Sometimes there is an extra layer of outline that shares the text colour
    # This parameter sets the weight to scrap from the text area
    nestingSuppression: 0
    # featureThreshold: minimum feature weight for a frame to be considered containing subtitles
    featureThreshold: 1.0
    # featureJumpThreshold: maximum feature difference between frames to be considered as a jump
    featureJumpThreshold: 0.1
    # featureJumpStddevThreshold: maximum standard deviation of feature difference between frames to be considered the same
    featureJumpStddevThreshold: 0.005
    # debugLevel: setting it to 1-3 will show different intermediate step images to the debug window
    debugLevel: 0
  
  yukkurilocaltrain:
    dialogRect: [0.00, 1.00, 0.75, 1.00]
    fastMode: false
    textWeightMin: 5
    textWeightMax: 25
    textHSVRanges:
      - [0, 200, 128, 30, 255, 255]
      - [170, 200, 128, 180, 255, 255]
      - [105, 100, 128, 135, 255, 255]
    outlineWeightMax: 5
    outlineHSVRanges:
      - [0, 0, 180, 180, 64, 255]
    boundCompensation: 4
    sobelThreshold: 192
    nestingSuppression: 9
    featureThreshold: 1.0
    featureJumpThreshold: 0.1
    featureJumpStddevThreshold: 0.005
    debugLevel: 0
  
  japantrafficlab:
    dialogRect: [0.00, 1.00, 0.75, 1.00]
    fastMode: false
    textWeightMin: 1
    textWeightMax: 19
    textHSVRanges:
      - [70, 180, 180, 100, 255, 255]
    outlineWeightMax: 15
    outlineHSVRanges:
      - [95, 180, 180, 140, 255, 255]
    boundCompensation: 4
    sobelThreshold: 100
    nestingSuppression: 0
    featureThreshold: 1.0
    featureJumpThreshold: 0.1
    featureJumpStddevThreshold: 0.005
    debugLevel: 0

  yorumeron-yukari:
    dialogRect: [0.00, 1.00, 0.80, 1.00]
    fastMode: false
    textWeightMin: 1
    textWeightMax: 15
    textHSVRanges:
      - [135, 0, 220, 155, 60, 255]
    outlineWeightMax: 9
    outlineHSVRanges:
      - [0, 0, 0, 180, 255, 25]
    boundCompensation: 2
    sobelThreshold: 192
    nestingSuppression: 0
    featureThreshold: 1.0
    featureJumpThreshold: 0.1
    featureJumpStddevThreshold: 0.005
    debugLevel: 0

  hotelzdm:
    dialogRect: [0.00, 1.00, 0.75, 1.00]
    fastMode: false
    textWeightMin: 5
    textWeightMax: 25
    textHSVRanges:
      - [0, 200, 100, 10, 255, 180]
      - [170, 200, 100, 180, 255, 180]
      - [25, 100, 100, 55, 200, 200]
    outlineWeightMax: 7
    outlineHSVRanges:
      - [0, 0, 200, 180, 32, 255]
    boundCompensation: 2
    sobelThreshold: 200
    nestingSuppression: 5
    featureThreshold: 1.0
    featureJumpThreshold: 0.1
    featureJumpStddevThreshold: 0.005
    debugLevel: 0

  zundahouse:
    dialogRect: [0.00, 1.00, 0.75, 1.00]
    fastMode: false
    textWeightMin: 1
    textWeightMax: 25
    textHSVRanges:
      - [155, 100, 200, 180, 200, 255]
      - [55, 150, 150, 75, 220, 220]
    outlineWeightMax: 15
    outlineHSVRanges:
      - [0, 0, 240, 180, 64, 255]
    boundCompensation: 4
    sobelThreshold: 230
    nestingSuppression: 23
    featureThreshold: 1.0
    featureJumpThreshold: 0.1
    featureJumpStddevThreshold: 0.005
    debugLevel: 0

# Box colour stat (general-purpose)
bcs:
  default:
    # dialogRect: rectangle of the detected dialog area
    dialogRect: [0.00, 1.00, 0.75, 1.00]
    # sobelThreshold: threshold for the Sobel edge detection
    sobelThreshold: 32
    # featureThreshold: minimum feature weight for a frame to be considered containing subtitles
    featureThreshold: 0.3
    # featureJumpThreshold: maximum feature difference between frames to be considered as a jump
    featureJumpThreshold: 0.1
    # featureJumpStddevThreshold: maximum standard deviation of feature difference between frames to be considered the same
    featureJumpStddevThreshold: 0.005
    # boxVerticalExpansion: vertical expansion ratio of the detected dialog box
    boxVerticalExpansion: 0.05
    # nonMajorBoxSuppressionMaxRatio: boxes with a ratio (over all boxes) lower than this will be suppressed
    nonMajorBoxSuppressionMaxRatio: 0.2
    # nonMajorBoxSuppressionMinRank: but at least this number of boxes will be kept
    nonMajorBoxSuppressionMinRank: 1
    # minCcAreaRatio: minimum ratio (over all connected components) to be considered text area candidate
    minCcAreaRatio: 0.0
    # maxCcAreaRatio: maximum ratio (over all connected components) to be considered text area candidate
    maxCcAreaRatio: 0.3
    # minCcFinalMean: minimum mean of connected component colour for a frame to be considered containing subtitles
    minCcFinalMean: 3.0
    # maxCcStddev: maximum standard deviation for a connected component to be considered consistent enough to be text
    maxCcStddev: 10
    # colourTolerance: distance from the selected mean colour to be filtered for text
    colourTolerance: 30
    # clusterThreshold: maximum distance threshold for hierarchical clustering
    clusterThreshold: 1
    # minColourAreaRatio: minimum ratio (over all connected components) to be further considered text area candidate
    minColourAreaRatio: 0.01
    # maxGreyscalePenalty: a colour that is right on the greyscale axis will be penalized by this factor for its tendency to be selected
    maxGreyscalePenalty: 0.70
    # iirPassDenoiseMinTime: minimum time (in milliseconds) for an interval to be kept
    iirPassDenoiseMinTime: 300
    # debugLevel: setting it to 1 will show the binary image of the text area
    debugLevel: 0

# Engines

# Engines decide how frames are sampled before they are processed by Strategies.
# This affects the performance and accuracy of the program.
# A given Strategy must implement a given Engine's required methods to work with it.
# Not all Engines are compatible with all Strategies.
# By default, all Strategies are compatible with the Framewise Engine.
# Currently, bcs and otl also support the Specilative Engine for faster processing.
# Each Engine has its own parameters that can be adjusted.

# engine: which frame sampling pipeline to use
# - framewise: samples every frame or every n-th frame
# - speculative: skips frames where subtitles do not change and performs binary search for change points
engine: speculative

framewise:
  # sampleInterval: in how many frames to sample a frame
  # If you have a high-resolution or high-fps video, you can increase this number to speed up the program.
  # However, this will reduce the accuracy of the subtitles.
  sampleInterval: 1
  # debug: whether to show debug information
  # Framewise Engine's debug mode will show a window displaying the processed frames with rectangles and FPIR parameters.
  # Some Strategies may also show intermediate steps in the debug window. 
  # Tap q to quit the debug window. The x button may not work. 
  debug: true
  # debugPyrDown: how many times to downscale the debug window
  # If the video's resolution is too high, the debug window may be too large to fit on the screen. 
  # You can downscale the debug window by this number to fit it on the screen.
  debugPyrDown: 0

speculative:
  # emptyGapForceCheck: in how many milliseconds to force check for existence of subtitles
  # If two frames were sampled where none of them had subtitles,
  # the program will force check the existence of subtitles in this number of milliseconds,
  # instead of considering there are no subtitles in the gap.
  # If this is set to 0, the program will not force check. 
  emptyGapForceCheck: 1000
  # debug: whether to show debug information
  # Speculative Engine's debug mode will print its decision of inserting/extending intervals after each frame.
  debug: false

# Extra Jobs

# Extra jobs are additional tasks that can be performed after the subtitles are extracted.
# Currently, the program supports only one experimental extra job: OCR.

# extraJobs: list of extra jobs to perform after subtitle extraction
extraJobs: []

ocr:
  suffix: "-ocr.txt"
  separator: "@"
  doPaddle: true
  paddleLang: "japan"
  doTesseract: true
  tesseractLang: "jpn"