kwzVideo.py

# =========================
# kwzVideo.py version 1.0.0
# =========================
# 
# Convert kwzs to mp4, using ffmpeg
# (no audio for now)
# 
# Usage:
# python kwzVideo.py <input.kwz> <output.mp4>

from sys import argv
from kwz import KWZParser, PALETTE
import subprocess as sp
import numpy as np

# convert palette to bytes
PALETTE = [bytes(color) for color in PALETTE]

if len(argv) != 3:
    print("\nUsage: python3 kwzVideo.py <input.kwz> <output.mp4>\n")
    exit(1)
else:
  infile = argv[1]
  outfile = argv[2]

  with open(infile, "rb") as kwz:
    parser = KWZParser(kwz)

    process = sp.Popen([
      "ffmpeg",
      # Hide the FFMPEG banner
      "-hide_banner",
      "-loglevel", "info",
      "-y",
      # Set the input frame size
      "-s", "320x240",
      # Set the frame rate
      "-r", str(parser.framerate),
      # Input format is raw, uncompressed frames
      "-f", "rawvideo",
      # thread_queue_size is the buffer size of the packets which are waiting to be encoded.
      # The default value is 8, but I was getting a "Thread message queue blocking; consider raising the thread_queue_size option"
      "-thread_queue_size", "600",
      # Input color format is rgb
      "-pix_fmt", "rgb24",
      # Read frame input from stdin
      "-i", "pipe:0",
      # Set output codecs and other settings to the most suitable values for HTML5 video playback
      "-vcodec", "libx264",
      "-pix_fmt", "yuv420p",
      # Rescale each frame, using nearest neighbor interpolation
      "-vf", "scale=%d:%d:flags=neighbor" % (320*2, 240*2),
      # Use all available threads for slight speed boost
      "-threads", "0",
      # Set compression preset
      "-preset", "medium",
      # Tune output for animation
      "-tune", "animation",
      # "-f", "mp4",
      outfile
    ], stdin=sp.PIPE)

    image = np.zeros((240, 320), dtype="V3")

    for frame_index in range(parser.frame_count):
      colors = parser.get_frame_palette(frame_index)
      # fill image with paper color
      image.fill(PALETTE[colors[0]])
      layers = parser.decode_frame(frame_index)
      # merge layers into image using numpy magic (aka -- masks)
      image[layers[2] == 1] = PALETTE[colors[5]]
      image[layers[2] == 2] = PALETTE[colors[6]]
      image[layers[1] == 1] = PALETTE[colors[3]]
      image[layers[1] == 2] = PALETTE[colors[4]]
      image[layers[0] == 1] = PALETTE[colors[1]]
      image[layers[0] == 2] = PALETTE[colors[2]]
      # convert image to bytes
      process.stdin.write(image.tobytes())

    # Close the stdin pipe and wait for completion
    process.stdin.close()
    process.wait()

    print("\nFinished conversion!\n")