osci-music-player/render.py

#!/usr/bin/env python3

# Render image to Oscilloscope XY vector audio
#
# https://pypi.org/project/svgpathtools/
# https://dood.al/oscilloscope/
#
# ----------------------------------------------------------------------------
# Copyright (c) 2024 Thomas Buck (thomas@xythobuz.de)
# Copyright (c) 2024 Philipp Schönberger (mail@phschoen.de)
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU General Public License for more details.
#
# See <http://www.gnu.org/licenses/>.
# ----------------------------------------------------------------------------

import sys
import math
import wave
import argparse
from svgpathtools import svg2paths

def rot_p(p_center, p,  angle_d):
    angle = math.radians(angle_d)
    ox = p_center[0]
    oy =  p_center[1]
    qx = ox + math.cos(angle) * (p[0] - ox) - math.sin(angle) * (p[1] - oy)
    qy = oy + math.sin(angle) * (p[0] - ox) + math.cos(angle) * (p[1] - oy)
    p = [qx, qy]
    return p

def read_image(filename, path_steps, volume_percent, angle_d):
    paths, attributes = svg2paths(filename)
    print("paths={}".format(len(paths)))

    p0 = [paths[0][0].start.real, paths[0][0].start.imag]
    p_min = [p0[0], p0[1]]
    p_max = [p0[0], p0[1]]

    for path in paths:
        print("segments={}".format(len(path)))

        # find center
        dist_min = float('inf')
        dist_max = 0
        p_prev = p_min
        for segment in path:
            p = [segment.end.real, segment.end.imag]
            for i in range(0, 2):
                if p[i] < p_min[i]:
                    p_min[i] = p[i]
                if p[i] > p_max[i]:
                    p_max[i] = p[i]

            dist_curr = math.sqrt((p[0] - p_prev[0]) ** 2 + (p[1] - p_prev[1]) ** 2)
            p_prev = p
            if dist_curr > dist_max:
                dist_max = dist_curr
            if dist_curr < dist_min:
                dist_min = dist_curr

    p_center = [ p_min[0] + (p_max[0] - p_min[0]) / 2, p_min[1] + (p_max[1] - p_min[1]) / 2 ]

    for path in paths:
        # TODO this is not nice.
        # Doing both range(0, 360, 5) and angle_d is redundant.
        # But also, it bakes in the assumption of the animation
        # that's specified in the Makefile.
        # Ideally the animation should be handled here in the
        # script instead, with proper parameters.

        # find min max for all rotations
        for segment in path:
            p_org = [segment.end.real, segment.end.imag]
            for a in range(0, 360, 5):
                p = rot_p(p_center, p_org , a)
                for i in range(0, 2):
                    if p[i] < p_min[i]:
                        p_min[i] = p[i]
                    if p[i] > p_max[i]:
                        p_max[i] = p[i]

        # find min max for this specific rotation
        for segment in path:
            p = [segment.end.real, segment.end.imag]
            p = rot_p(p_center, p, angle_d)
            for i in range(0, 2):
                if p[i] < p_min[i]:
                    p_min[i] = p[i]
                if p[i] > p_max[i]:
                    p_max[i] = p[i]

    print("min={} max={}".format(p_min, p_max))
    print("center={} ".format(p_center))
    print("dist min={} max={} ".format(dist_min, dist_max))

    data = bytearray()

    def add_point(p):
        for i in range(0, 2):
            v = p[i]
            v -= p_min[i]
            v /= p_max[i] - p_min[i]
            if i == 1:
                v = 1 - v
            c = int((v * 2 - 1) * (32767 / 100 * volume_percent))
            data.extend(c.to_bytes(2, byteorder="little", signed=True))

    def interpolate(p1, p2, step):
        p = []
        for i in range(0, 2):
            diff = p2[i] - p1[i]
            v = p1[i] + diff * step
            p.append(v)
        return p

    def add_segment(p1, p2, f):
        p = interpolate(p1, p2, f)
        add_point(p)

    def add_path(p1, p2):
        l = math.sqrt((p2[0] - p1[0]) ** 2 + (p2[1] - p1[1]) ** 2)
        ps = max(1, int(path_steps * l))
        for step in range(0, ps):
            add_segment(p1, p2, step / ps)

    for path in paths:
        p = [path[0].start.real, path[0].start.imag]
        p = rot_p(p_center, p , angle_d)
        points = [p]
        for segment in path:
            p = [segment.end.real, segment.end.imag]
            p = rot_p(p_center, p , angle_d)
            points.append(p)

        # walk path forwards
        for n in range(0, len(points) - 1):
            add_path(points[n], points[n + 1])
        add_point(points[len(points) - 1])

        # walk path backwards
        for n in range(len(points) - 2, -1, -1):
            add_path(points[n + 1], points[n])
        add_point(points[0])

    return data

def write_waveform(data, filename, samplerate):
    with wave.open(filename, "w") as f:
        f.setnchannels(2)
        f.setsampwidth(2)
        f.setframerate(samplerate)
        f.writeframes(data)

def main():
    parser = argparse.ArgumentParser(
        prog=sys.argv[0],
        description='Render SVG path to vector XY audio file',
        epilog='Made by Thomas Buck (thomas@xythobuz.de) and Philipp Schönberger (mail@phschoen.de). Licensed as GPLv3.')

    parser.add_argument("input", help="Input SVG image file path.")
    parser.add_argument("-o", "--output", dest="output", default="out.wav",
                        help="Output wav sound file path. Defaults to 'out.wav'.")
    parser.add_argument("-t", "--time", dest="time", default=5.0, type=float,
                        help="Length of sound file in seconds. Defaults to 5s.")
    parser.add_argument("-s", "--samplerate", dest="samplerate", default=44100, type=int,
                        help="Samplerate of output file in Hz. Defaults to 44.1kHz.")
    parser.add_argument("-v", "--volume", dest="volume", default=100.0, type=float,
                        help="Volume of output file in percent. Defaults to 100%%.")
    parser.add_argument("-i", "--interpolate", dest="interpolate", default=3, type=int,
                        help="Steps on interpolated paths. Defaults to 3.")
    parser.add_argument("-r", "--rotate", dest="angle_d", default=0.0, type=float,
                        help="Angle to rotate image, in degrees. Defaults to 0 deg.")

    args = parser.parse_args()
    print(args)

    wave = read_image(args.input, args.interpolate, args.volume, args.angle_d)

    samplecount = int(len(wave) / 2 / 2) # stereo, int16
    drawrate = args.samplerate / samplecount
    drawcount = drawrate * args.time
    print("len={} samples={} drawrate={:.2f} count={:.2f}".format(len(wave), samplecount, drawrate, drawcount))

    data = bytearray()
    for n in range(0, int(drawcount)):
        data.extend(wave)
    print("len={}".format(len(data)))

    write_waveform(bytes(data), args.output, args.samplerate)

if __name__ == "__main__":
    main()