// (Triforce, 11-19) The title list's text as the game draws it round its entries (PaCommonBtnTitle_00, the Title archive's // layout): not chosen, T_BtnSh_00, the text in black (the layout's colour 3) 2 units under it, spread soft into a dark edge; // chosen, P_BtnGlow_00, the capture C_Btn_00 of the text (T_Btn_00) blurred, under the white text (T_Btn_02) in blue. Their // materials are the font's effect (an extension of 28 bytes where the plain text's has 24: T_BtnSh_00's 70 and 30, // T_Btn_00's 40), which the overlay cannot read as the game does; ImGui has no blur either. So the text's own shape is // drawn again round it, ring after ring, each copy a little seen (the copies' alpha falls off as a Gaussian's), and the rings // were fitted to the game's own entries on the lab's shots (runs/port052, 720 picture lines: 3 units a pixel) with the // text's mask taken from the game's own "Continue": the darkness of New Game's and Options' edges at 1 to 8 pixels from their // strokes (0.47 at 2, 0.27 at 3, 0.13 at 4, 0.05 at 5) and the blue of the chosen Continue's and Options' glow (0.93 at // 1, 0.67 at 3, 0.40 at 5, 0.18 at 7, 0.01 at 9). The entry had one shadow copy at 150 (no edge to be seen) and four // rings out to 15 units (gone at 2 pixels): the lab's "much weaker dark shadow ... its glow is fainter". // ((Triforce, 11-20) The text not chosen fitted again. The lab's second check (shots052b): "paler and greyer than New Game // over the bright wheat, its dark edge weaker" in the 1720x720 and 1600x900 windows (the darkest 2% of pixels round it 34-41 // of 255, New Game's 19-21), close in the 3440x1440 fullscreen. The fit above took a stroke for a half plane, where a ring // copy far out still covers the place; beside a thin stroke it covers it only where the stroke is, so the edge drawn was // weaker near the strokes and broader than fitted. And the game's text is bolder than ImGui's Cetra Display: its strokes 6 // units wide in the fullscreen and both windows, ours 4.5; at 720 lines that is 1 whole pixel with grey ones beside it where // the game's have 2. It sits on a near black shadow tight along the strokes (the darkest pixels beside New Game 0.93-0.96 // dark in the shots, ours 0.75-0.80 at every size). At 1440 lines the 1.5 units are a pixel of 3 and the text passed; at 720 // or 674 they are half of its whole pixels: the greyer text. Fitted now with the game's own font drawn by ImGui as the game // draws it, in a model of the drawing (the copies, their 8-bit alpha, and the shots' HDR capture, where a black copy of alpha // a darkens 1-(1-a)^2, read from the old Mods in the same shots), to New Game's and Options' edges in the fullscreen's and // both windows' shots at once: the text drawn twice over itself (kTextWeight), a hard shadow (the core: the text in black // 2.5 units down, 16 copies of 0.45, so that its soft edges fill out as the game's outlined T_BtnSh_00's) and the soft edge // round both (the shadow). The distance from New Game's edge at 1440, 720 and 674 lines: 0.28, 0.27, 0.37 before; 0.06, 0.03, // 0.07 now. The copies are put at their offsets from the text's whole-pixel place, not each snapped to a pixel (vk_menu.cpp // draw_title_mods): the core's 2.5 units are 0.83 of a pixel at 720 lines, cut to 0 or 1 by where the entry happened to be. // ((Triforce, 11-20, lab check 3, shots052c) Fitted again, per output, in the five shapes the lab ran: fullscreen 3440x1440, // the 1720x720 window, the 1600x900 window (a 21:9 picture of 674 lines), the 1680x1050 window with a 16:10 picture (the // layouts squeezed across by 0.9) and the 1720x720 window with HDR off. What the lab saw: the strokes thinner than New // Game's (mean 3.5 pixels across for 4.3-4.6 at 1440 lines, 1.5 for 2.0 at 674), the hard shadow too dark and crisp at 720 // (its darkest 2% 0.0-0.9 of 255, New Game's 1.2-16.6), and with HDR off the edge a third of the game's (0.09-0.31 dark 6 // units out for 0.41-0.53). The text drawn twice in place has 2 copies' coverage at every size; the game's text, drawn at // 1440 lines and scaled to the window, is bolder in units at 1440 than at 720. So the weight is two copies in place and // two more either side across (Weight: their offset and alpha by the text's size in pixels: 0.87 units out at 0.88 of // alpha at 1440 lines, 0.76 at 0.43 at 720, 0.41 of alpha at 674), the hard shadow softer at 720, and with HDR off // (kLookSdr) every black copy as dark in 8-bit code values as the HDR picture shows it (the overlay blends there in PQ // code values, where a black copy of alpha a darkens about 1 - (1 - a)^2; in SDR about a), the text no wider. // No ImGui: tests/test_mods_title_entry.py compiles it alone. #pragma once #include namespace text_spread { struct Spread { float a; // (a copy's alpha at the middle, of 1) float sigma; // (units: the copies' alpha exp(-(r / sigma)^power)) float power; float r0, dr; // (units: the first ring and the step) int rings; float centre; // (a copy under the text itself, its alpha) float down; // (units: the whole of it under the text by this) int repeat = 1; // ((11-20) the copy under the text drawn this many times) }; /// P_BtnGlow_00's: the glow's blue (kTitleGlow) constexpr Spread kGlow{18.85f / 255.0f, 16.0f, 1.85f, 3.0f, 3.0f, 10, 1.0f, 0.0f}; /// ((11-20) the chosen text drawn this many times over itself: the game's strokes' weight (ImGui's Cetra Display is 1.5 units /// thinner; a pixel the text covers c of shows 1 - (1 - c)^2). (Lab check 3: the text not chosen has its own, Look's weight) constexpr int kTextWeight = 2; /// (the alpha of each of those draws for the text at alpha a: a where the text is whole) inline float weight_alpha(float a, int weight = kTextWeight) { return a >= 1 ? 1.0f : a <= 0 ? 0.0f : 1.0f - std::pow(1.0f - a, 1.0f / (float)weight); } inline float ring_radius(const Spread& s, int i) { return s.r0 + s.dr * (float)i; } inline float ring_alpha(const Spread& s, int i) { return s.a * std::exp(-std::pow(ring_radius(s, i) / s.sigma, s.power)); } /// (the copies of a ring: about one every 3 units round it, 8 at least) inline int ring_copies(const Spread& s, int i) { const int n = (int)std::ceil(2.0f * 3.14159265f * ring_radius(s, i) / 3.0f); return n < 8 ? 8 : n; } /// ((Triforce, 11-19) A ring as it is drawn: its copies' alpha a whole step of 255 (an 8-bit colour's), and as many copies as /// keep the ring's coverage (copies x alpha) the fit's. Its alpha is only 1 to 19 of 255, so cut to a whole step it lost a /// large share (the outer glow rings to 0, not drawn: the lab's weak edge and faint glow again), and the more while the /// entry fades in (amount under 1). amount: the entry's fade times the colour's own alpha (of 1) struct Drawn { int alpha; // (of 255; 0 none drawn) int copies; }; inline Drawn ring_drawn(const Spread& s, int i, float amount) { const float a = ring_alpha(s, i) * amount * 255.0f; const int n = ring_copies(s, i); if (!(a > 0)) return {0, 0}; int q = (int)std::lround(a); q = q < 1 ? 1 : q > 255 ? 255 : q; return {q, (int)std::lround((float)n * a / (float)q)}; } /// (the copy under the text itself: its alpha rounded to a whole step; drawn s.repeat times) /// ((Triforce, 11-20) drawn more than once (a Look's core), the copies' darkness together is amount times theirs at full /// strength, as the text's own alpha is amount (weight_alpha): each copy 1 - (1 - amount C)^(1 / repeat), C = 1 - (1 - a)^repeat /// with a the full copy's whole step. A review of cb644b1: each at amount x 0.45 stacked 16 times to 1 - (1 - 0.45 amount)^16, /// 0.98 black under the text at 0.5, 0.86 at 0.25: "Mods" a near black word as the list rose, the core left after the text /// as it went, and the blue glow darkened through UnSelect; the game's T_BtnSh_00 fades with its pane) inline int centre_drawn(const Spread& s, float amount) { long q = std::lround(s.centre * amount * 255.0f); if (s.repeat > 1 && amount > 0 && amount < 1) { const float full = (float)std::lround(s.centre * 255.0f) / 255.0f; const float dark = 1.0f - std::pow(1.0f - full, (float)s.repeat); q = std::lround((1.0f - std::pow(1.0f - amount * dark, 1.0f / (float)s.repeat)) * 255.0f); } return q < 0 ? 0 : q > 255 ? 255 : (int)q; } /// ((11-20) every copy of a spread as draw_title_mods draws it: f(dx, dy, alpha of 255), dx and dy in units from the text's /// place (the caller takes them to pixels), the copies under the text first) template inline void each_copy(const Spread& s, float amount, F&& f) { if (const int a = s.centre > 0 ? centre_drawn(s, amount) : 0) for (int n = 0; n < s.repeat; n++) f(0.0f, s.down, a); for (int i = 0; i < s.rings; i++) { const Drawn dr = ring_drawn(s, i, amount); if (!dr.alpha || !dr.copies) continue; const float r = ring_radius(s, i); for (int j = 0; j < dr.copies; j++) { const float ang = 6.2831853f * (float)j / (float)dr.copies; f(std::cos(ang) * r, s.down + std::sin(ang) * r, dr.alpha); } } } /// ((Triforce, 11-20, lab check 3) The text not chosen as its weight: `same` copies at its own place, `wide` copies round it /// (2: either side across) `widen` units out and `wide_a` of alpha, both by the text's size: lines, the picture's lines /// times the list's scale (2160 units). (widen x (lines / 1080)^widen_pow, wide_a x (lines / 720)^wide_pow, 1 at most) struct Weight { int same; int wide; float widen, widen_pow; float wide_a, wide_pow; }; inline float weight_widen(const Weight& w, float lines) { return lines > 0 ? w.widen * std::pow(lines / 1080.0f, w.widen_pow) : 0.0f; } inline float weight_wide_a(const Weight& w, float lines) { const float a = lines > 0 ? w.wide_a * std::pow(lines / 720.0f, w.wide_pow) : 0.0f; return a < 0 ? 0.0f : a > 1 ? 1.0f : a; } /// (every copy of the text as draw_title_mods draws it at amount (its fade, of 1): f(dx, dy, alpha of 255), dx and dy in /// units. Each copy's alpha the full one's times a scale g, found so that where the text is whole the copies together are /// amount opaque, as the text once at alpha amount ((1 - g)^same (1 - g wide_a)^wide = 1 - amount): no brighter step /// through the fades) template inline void each_weight_copy(const Weight& w, float lines, float amount, F&& f) { if (!(amount > 0)) return; const float wa = w.wide ? weight_wide_a(w, lines) : 0.0f, r = weight_widen(w, lines); float g = 1; if (amount < 1) { float lo = 0, hi = 1; for (int i = 0; i < 24; i++) { g = (lo + hi) * 0.5f; const float through = std::pow(1 - g, (float)w.same) * std::pow(1 - g * wa, (float)w.wide); (1 - through < amount ? lo : hi) = g; } g = (lo + hi) * 0.5f; } const int same_a = (int)std::lround(g * 255.0f), wide_q = (int)std::lround(g * wa * 255.0f); for (int i = 0; i < w.same; i++) if (same_a > 0) f(0.0f, 0.0f, same_a > 255 ? 255 : same_a); for (int j = 0; j < w.wide && wide_q > 0; j++) { const float ang = 6.2831853f * (float)j / (float)w.wide; f(std::cos(ang) * r, std::sin(ang) * r, wide_q > 255 ? 255 : wide_q); } } /// ((Triforce, 11-20, lab check 3) the text not chosen, per output: T_BtnSh_00's hard shadow (core: the text in black under /// it, repeat times, and a ring of 8 faint copies a third of a unit round that) and its soft edge (shadow), drawn in that order /// before the text's weight copies struct Look { Spread core; Spread shadow; Weight weight; }; /// (HDR10 window images: the overlay in PQ code values. Fitted to New Game, Options and Continue at 1440, 1050 (16:10, /// squeezed across), 720 and 674 lines at once) constexpr Look kLookHdr{ {15.9f / 255.0f, 1000.0f, 1.0f, 0.37f, 0.0f, 1, 107.6f / 255.0f, 2.55f, 14}, {10.0f / 255.0f, 11.19f, 1.21f, 1.1f, 2.71f, 7, 0.0f, 4.0f}, {2, 2, 0.82f, 0.2f, 0.434f, 1.02f}, }; /// (SDR window images: 8-bit code values. Fitted at 720 lines; the copies' alpha about 1 - (1 - a)^2 of HDR's, the text no /// wider across (the game's own strokes are thinner in SDR: 1.74 pixels at 720 lines, 2.06 in HDR)) constexpr Look kLookSdr{ {0.0f, 1000.0f, 1.0f, 0.37f, 0.0f, 0, 170.5f / 255.0f, 2.55f, 14}, {21.1f / 255.0f, 11.19f, 1.21f, 1.1f, 2.71f, 7, 0.0f, 4.0f}, {2, 0, 0.0f, 0.0f, 0.0f, 0.0f}, }; inline const Look& look_of(bool hdr) { return hdr ? kLookHdr : kLookSdr; } } // namespace text_spread