arisusay 1.0.1

An ASCII teabag player for Tendou Aris — like momoisay, but she teabags. Homage to 청세치/세치혀.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163

#!/usr/bin/env python3
"""Generate Braille ASCII frames of Tendou Aris teabagging.

Like momoisay, Aris is rendered ALONE (background removed) as filled Braille dots.
A UNION bounding box per source keeps her from being clipped or jittering; a dot
is filled where the pixel is opaque AND dark. Detached blobs such as her halo
bracket are kept (all interior blobs survive the chroma key) and are filled as a
silhouette so the pale halo stays visible at Braille resolution.

Sources (in refs/):
  * aris2.gif    — transparent-background teabag loop (its halo reads at Braille res)
  * yt_src.webm  — green-screen clip around 1:05-1:27 (keyframe slop shifts it a few
        seconds earlier); motion_a = 8.5-13s = the single-Aris bob after the clones
  * yt_src3.mp4  — green-screen clip around 1:14-1:30;
        motion_b = 5.75-11.3s = the BOTH-HANDS-UP bob (the 1:20-1:24 bit): first
        three and last two checkHU2 cells dropped, ends before a second Aris
        walks in

Reproducible: `python3 tools/gen_frames.py`   (needs: Pillow, numpy, scipy, ffmpeg)

Emits frames/<name>.txt (frames joined by form-feed \\x0c) + frames/static.txt.
"""
import os
import subprocess
import tempfile

import numpy as np
from PIL import Image
from scipy import ndimage

WCELLS = 34
DARK = 0.55          # luminance below this (over white) -> dot
MIN_BLOB = 60        # keep chroma-keyed blobs at least this many px (halo ~hundreds)
PAD = 4
SEP = "\x0c"

# name -> dict describing how to get RGBA frames + alpha/minor masks
SOURCES = {
    "aris":     {"kind": "gif",   "path": "refs/aris2.gif"},
    "motion_a": {"kind": "green", "path": "refs/yt_src.webm",  "ss": 8.5,  "to": 13,   "fps": 12},
    "motion_b": {"kind": "green", "path": "refs/yt_src3.mp4",  "ss": 5.75, "to": 11.3, "fps": 12},
}
STATIC_FROM = "aris"

_DOTS = [(0, 0, 0), (1, 0, 1), (2, 0, 2), (0, 1, 3),
         (1, 1, 4), (2, 1, 5), (3, 0, 6), (3, 1, 7)]


def green_alpha(rgb):
    """Chroma-key. Returns (alpha, minor) as uint8 masks.

    alpha: every non-green interior blob >= MIN_BLOB px (body, halo, ...).
    minor: those blobs except the largest (the halo etc.) — force-filled later
    so pale detached parts stay visible.
    """
    r, g, b = rgb[..., 0].astype(int), rgb[..., 1].astype(int), rgb[..., 2].astype(int)
    green = (g > 100) & (g > r + 30) & (g > b + 30)
    nong = ~green
    lbl, _ = ndimage.label(nong)  # type: ignore[misc]
    border = set(lbl[0, :]) | set(lbl[-1, :]) | set(lbl[:, 0]) | set(lbl[:, -1])
    border.discard(0)
    interior = nong & ~np.isin(lbl, list(border))
    l2, n2 = ndimage.label(interior)  # type: ignore[misc]
    alpha = np.zeros(rgb.shape[:2], bool)
    minor = np.zeros(rgb.shape[:2], bool)
    if n2 > 0:
        sizes = ndimage.sum(np.ones_like(l2), l2, range(1, n2 + 1))
        main = int(np.argmax(sizes)) + 1
        for i, sz in enumerate(sizes, start=1):
            if sz >= MIN_BLOB:
                blob = l2 == i
                alpha |= blob
                if i != main:
                    minor |= blob
    return (alpha * 255).astype("uint8"), (minor * 255).astype("uint8")


def load_source(spec):
    """Return a list of (rgba_image, alpha_uint8, minor_uint8) per frame."""
    if spec["kind"] == "gif":
        im = Image.open(spec["path"])
        out = []
        for i in range(getattr(im, "n_frames", 1)):
            im.seek(i)
            f = im.convert("RGBA")
            a = np.asarray(f)[..., 3]
            out.append((f, a, np.zeros_like(a)))
        return out
    # green-screen video segment
    with tempfile.TemporaryDirectory() as td:
        subprocess.run(["ffmpeg", "-y", "-ss", str(spec["ss"]), "-to", str(spec["to"]),
                        "-i", spec["path"], "-vf", f"fps={spec['fps']}", os.path.join(td, "f%04d.png")],
                       check=True, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
        out = []
        for name in sorted(os.listdir(td)):
            im = Image.open(os.path.join(td, name)).convert("RGB")
            rgb = np.asarray(im)
            alpha, minor = green_alpha(rgb)
            # The halo bracket in this source is pale teal — too bright for the
            # dark-fill — so force-fill teal-ish opaque pixels alongside the
            # detached blobs. (Gif sources are untouched.)
            r, g, b = rgb[..., 0].astype(int), rgb[..., 1].astype(int), rgb[..., 2].astype(int)
            teal = (b > r + 25) & (g > r + 15) & (alpha > 0)
            minor = np.maximum(minor, (teal * 255).astype("uint8"))
            out.append((im.convert("RGBA"), alpha, minor))
        return out


def union_box(frames):
    x0 = y0 = 1 << 30
    x1 = y1 = 0
    for _, a, _ in frames:
        ys, xs = np.where(a > 128)
        if len(xs):
            x0, x1 = min(x0, int(xs.min())), max(x1, int(xs.max()))
            y0, y1 = min(y0, int(ys.min())), max(y1, int(ys.max()))
    w, h = frames[0][0].size
    return (max(0, x0 - PAD), max(0, y0 - PAD), min(w, x1 + PAD), min(h, y1 + PAD))


def to_braille(rgba, alpha, minor, box):
    im = rgba.crop(box)
    a = np.asarray(Image.fromarray(alpha).crop(box), np.float32) / 255.0
    mn = np.asarray(Image.fromarray(minor).crop(box), np.float32) / 255.0
    rgb = np.asarray(im.convert("RGBA"), np.float32)[..., :3] / 255.0
    white = rgb * a[..., None] + (1 - a[..., None])
    w, h = im.size
    wpx = WCELLS * 2
    hpx = max(4, round(wpx * h / w))
    lum = Image.fromarray((white * 255).astype("uint8")).convert("L").resize((wpx, hpx), Image.Resampling.LANCZOS)
    opa = Image.fromarray((a * 255).astype("uint8")).resize((wpx, hpx), Image.Resampling.LANCZOS)
    mno = Image.fromarray((mn * 255).astype("uint8")).resize((wpx, hpx), Image.Resampling.LANCZOS)
    g = np.asarray(lum, np.float32) / 255.0
    op = np.asarray(opa, np.float32) / 255.0
    mr = np.asarray(mno, np.float32) / 255.0
    mask = ((op > 0.5) & (g < DARK)) | (mr > 0.5)

    H = (hpx + 3) // 4 * 4
    W = (wpx + 1) // 2 * 2
    m = np.zeros((H, W), bool)
    m[:hpx, :wpx] = mask
    lines = []
    for by in range(0, H, 4):
        row = []
        for bx in range(0, W, 2):
            bits = sum((1 << bit) for r, c, bit in _DOTS if m[by + r, bx + c])
            row.append(chr(0x2800 + bits))
        lines.append("".join(row))
    return "\n".join(lines)


if __name__ == "__main__":
    os.makedirs("frames", exist_ok=True)
    first = {}
    for name, spec in SOURCES.items():
        frames = load_source(spec)
        box = union_box(frames)
        arts = [to_braille(f, a, mn, box) for f, a, mn in frames]
        open(f"frames/{name}.txt", "w").write(SEP.join(arts))
        first[name] = arts[0]
        print(f"{name}: {len(arts)} frames  box={box}")
    open("frames/static.txt", "w").write(first[STATIC_FROM])
    print("static: 1 frame")