perhaps-aksh · Oct 13, 2026 · Oct 13, 2026 · Oct 13, 2026
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -2,7 +2,14 @@
 
 All notable changes, newest first. The project follows [Conventional Commits](https://www.conventionalcommits.org/); see `git log` for every commit.
 
-## [Unreleased]: public release prep
+## 2026-10-13: Painted art (Day 4)
+- The Last Train is painted: 28 world assets (sky to signal box, the train, Detective Grace, every clue object), placed to line up with what's painted.
+- Clue clocks keep exact procedural hands over the painted dials (new manifest `overlay`).
+- Evidence cards show painted close-ups; the speech bubble shows painted portraits, with a "pressed" expression when shown evidence.
+- Home and library key art; case cards with painted headers.
+- Fix: a lightning flash cut short no longer leaves a grey veil.
+
+## 2026-10-13: Public release
 - **Brand:** the app is now "Room Riddle", with a launcher icon on the TV and a favicon and web manifest on the phone.
 - **Docs:** showcase README with screenshots; technical reference moved to `docs/system-design.md`; contributing guide, code of conduct, security policy, credits, `.env.example`, CI and issue/PR templates.
 

diff --git a/CREDITS.md b/CREDITS.md
@@ -4,7 +4,7 @@
 
 ## Art
 - App icon, banner and concept art: AI-assisted illustration, art-directed and edited by the author. Released with the project under the MIT licence.
-- TV scenes in the current build use code-drawn placeholder shapes; painted plates and character art are being added through the asset system.
+- The Last Train's scenes, objects, characters, clue close-ups and the menu key art are AI-assisted paintings, art-directed, cut out and edited by the author (`apps/tv/scripts/art/` sizes them for the TV). The other cases still use code-drawn placeholder shapes.
 - Screenshots in `docs/images` were captured from the Vega Virtual Device and a phone browser.
 
 ## Audio

diff --git a/README.md b/README.md
@@ -34,7 +34,7 @@
 
 <p align="center">
   <img src="docs/images/hero-tv-phone.jpg" alt="A suspect answers on the TV while the player questions him from a phone" width="100%" />
-  <br/><sub>Asking Inspector Voss a question from a phone. He answers on the TV, in character and in his own voice.</sub>
+  <br/><sub>Asking Inspector Voss a question from a phone. He answers on the TV, in character, in his own voice, with his painted portrait.</sub>
 </p>
 
 1. **Open a room on the TV.** It shows a 4-digit code and a QR code.
@@ -48,8 +48,21 @@
 |:---:|:---:|
 | <img src="docs/images/tv-home.jpg" alt="Home screen" /> | <img src="docs/images/tv-library.jpg" alt="The case library" /> |
 | **Home:** create or join a room with the remote | **Library:** five complete cases |
-| <img src="docs/images/tv-lobby.jpg" alt="Room lobby with QR code" /> | <img src="docs/images/tv-station.jpg" alt="The 2.5D station scene" /> |
-| **Lobby:** scan the QR code, see who's in | **The world:** a layered 2.5D scene you explore with the remote |
+| <img src="docs/images/tv-lobby.jpg" alt="Room lobby with QR code" /> | <img src="docs/images/tv-arrival.jpg" alt="The Northern Star arriving at Harrow Cross" /> |
+| **Lobby:** scan the QR code, see who's in | **Arrival:** the Northern Star pulls in through the rain |
+
+### Inside The Last Train
+
+Painted 2.5D scenes: depth layers move with the camera, rain and steam play over the art, and every clue object sits where it's painted.
+
+| | |
+|:---:|:---:|
+| <img src="docs/images/tv-station.jpg" alt="Platform two, explorable with the remote" /> | <img src="docs/images/tv-clue.jpg" alt="Evidence card with a painted close-up" /> |
+| **Platform two:** focus objects with the remote | **Evidence:** every clue has its painted close-up |
+| <img src="docs/images/tv-vestibule.jpg" alt="Detective Grace in the carriage doorway" /> | <img src="docs/images/tv-ticket.jpg" alt="The torn ticket beside the train" /> |
+| **Detective Grace** narrates and reacts to what you find | **On the platform:** a torn ticket in a puddle |
+| <img src="docs/images/tv-corridor.jpg" alt="The sleeper corridor of carriage C" /> | <img src="docs/images/tv-signal-box.jpg" alt="The signal box at two minutes to midnight" /> |
+| **Carriage C:** the sleeper corridor | **The signal box:** two minutes to midnight |
 
 ### Your phone is your notebook
 
@@ -58,8 +71,6 @@
   <br/><sub>Join in seconds · examine evidence · question suspects · solve puzzles with three-step hints</sub>
 </p>
 
-> **Art note:** the TV scenes in these screenshots use the built-in placeholder art. The painted art pass (illustrated plates, character portraits, video loops) goes in through the asset system without code changes; see [Status](#-status).
-
 ---
 
 ## ✦ The cases
@@ -141,7 +152,7 @@ npm run serve          # build the phone app and start the server on :4000
 **TV app** (inside Ubuntu or WSL2 with the Vega SDK):
 
 ```bash
-bash apps/tv/scripts/wsl-check.sh      # typecheck + lint + 75 Jest tests
+bash apps/tv/scripts/wsl-check.sh      # typecheck + lint + 80 Jest tests
 bash apps/tv/scripts/wsl-build.sh      # build the .vpkg
 bash apps/tv/scripts/wsl-install.sh    # install and launch on the Vega Virtual Device
 ```
@@ -214,7 +225,7 @@ The full technical reference is in **[docs/system-design.md](docs/system-design.
 
 ## ✦ Quality
 
-- **About 320 automated tests:** 58 engine, 22 content, 52 scene, 19 client, 4 UI tokens, 78 server, 11 phone and 75 TV.
+- **About 325 automated tests:** 58 engine, 22 content, 52 scene, 19 client, 4 UI tokens, 78 server, 11 phone and 80 TV.
 - **Every case is played to the end automatically.** Content validation fails the build on unreachable scenes, broken references, puzzles that need clues from later scenes, and spoilers that could never unlock.
 - **Spoiler-attack tests** run against the AI guard and the prompt builder.
 - **Scripted device QA:** the Vega Virtual Device is driven with remote key presses and screenshots, with an audio trace sink.
@@ -231,7 +242,8 @@ The full technical reference is in **[docs/system-design.md](docs/system-design.
 | Phone app: Scene, Evidence and Question tabs | ✅ Done, tested |
 | TV app: screens, lobby, settings, library, 2.5D world, audio | ✅ Done, verified on the Vega Virtual Device |
 | AI suspect questioning with voices on the TV | ✅ Done, verified on the Vega Virtual Device |
-| Painted art, video loops and final audio | 🎨 In progress (assets in production) |
+| Painted art for The Last Train, portraits, clue close-ups, menu key art | ✅ Done, verified on the Vega Virtual Device |
+| Art for the other four cases, video loops, final audio | 🎨 In progress |
 | UI redesign to the final art direction | 🎨 In progress |
 | AI hints and narrator | 🔜 Next |
 | Validation on Fire TV Stick hardware | 🔜 Next |

diff --git a/apps/tv/jest.config.json b/apps/tv/jest.config.json
@@ -16,7 +16,7 @@
     "^~/(.*)$": "<rootDir>/src/$1",
     "^@amazon-devices/react-native-async-storage__async-storage$": "<rootDir>/node_modules/@amazon-devices/react-native-async-storage__async-storage/jest/async-storage-mock",
     "^@amazon-devices/react-native-w3cmedia$": "<rootDir>/test/support/w3cmedia.ts",
-    "\\.(mp3|m4a|wav)$": "<rootDir>/test/support/fileStub.js"
+    "\\.(mp3|m4a|wav|png|jpg)$": "<rootDir>/test/support/fileStub.js"
   },
   "modulePaths": [
     "<rootDir>/src"

diff --git a/apps/tv/scripts/art/prepare_last_train.py b/apps/tv/scripts/art/prepare_last_train.py
@@ -0,0 +1,247 @@
+#!/usr/bin/env python3
+"""
+Prepare The Last Train's painted art for the TV app.
+
+Reads the raw art (PNG, transparent where needed) from SRC and writes files
+sized exactly to each asset's manifest box into
+apps/tv/src/assets/mysteries/the-last-train/, so the TV can draw them with
+resizeMode "stretch" without distortion:
+
+  cover   fill the box, crop the overflow (backgrounds)
+  contain fit inside the box, pad with transparency (objects, characters)
+  tile    repeat horizontally, mirroring every other copy to hide seams
+
+Objects are stored at 2x their box (they are zoomed into when inspected);
+backgrounds at 1x. Opaque layers are JPEG, anything with alpha is PNG.
+
+It also records where the procedural clock hands go on each painted dial
+(measured by eye, see SPEC) (the hands show the clue times, so they stay procedural).
+
+Usage: python prepare_last_train.py <SRC art-inbox dir> [--only id,id]
+Requires Pillow, numpy, scipy.
+"""
+import json
+import sys
+from pathlib import Path
+
+import numpy as np
+from PIL import Image, ImageDraw, ImageFilter
+from scipy import ndimage
+
+REPO = Path(__file__).resolve().parents[4]
+OUT = REPO / 'apps/tv/src/assets/mysteries/the-last-train'
+
+# id: (source relative to SRC, box w, box h, mode, options)
+SPEC = {
+    'sky': ('last-train/last-train.sky.png', 2600, 1200, 'cover', {}),
+    'city': ('last-train/last-train.city.png', 3400, 360, 'band', {'from': 0.52, 'fade': 0.45}),
+    'station.background': ('last-train/last-train.station.background.png', 1700, 760, 'cover', {'fade': 0.22}),
+    'station.platform': ('last-train/last-train.station.platform.png', 5600, 320, 'tile', {}),
+    'tracks': ('last-train/last-train.tracks.png', 5600, 90, 'tile', {}),
+    'station.lights': ('last-train/last-train.station.lights.png', 90, 560, 'contain', {'align': 'bottom'}),
+    'foreground.pole': ('last-train/last-train.foreground.pole.png', 110, 1200, 'contain', {'align': 'bottom', 'stretch': True}),
+    'bench': ('last-train/last-train.bench.png', 320, 120, 'contain', {'align': 'bottom'}),
+    'train': ('last-train/last-train.train.png', 2600, 380, 'contain', {'align': 'bottom', 'stretch': True}),
+    'train.interior': ('last-train/last-train.train.interior.png', 2400, 1180, 'cover', {}),
+    'train.vestibule': ('last-train/last-train.train.vestibule.png', 2200, 1080, 'cover', {}),
+    'signal-box.interior': ('last-train/last-train.signal-box.interior.png', 2300, 1180, 'cover', {}),
+    'detective': ('shared/shared.grace.full.png', 240, 560, 'contain', {'align': 'bottom', 'scale': 2}),
+    'clock': ('last-train/last-train.clock.png', 200, 200, 'contain', {'scale': 2, 'face': {'cx': 101, 'cy': 101, 'd': 118}}),
+    'notice': ('last-train/last-train.notice.png', 180, 120, 'contain', {'scale': 2}),
+    'waiting-room-clock': ('last-train/last-train.waiting-room-clock.png', 90, 90, 'disc', {'scale': 3, 'at': (0.502, 0.445, 0.27), 'face': {'cx': 45, 'cy': 45, 'd': 62}}),
+    'suitcase': ('last-train/last-train.suitcase.png', 170, 120, 'contain', {'align': 'bottom', 'scale': 2}),
+    'newspaper': ('last-train/last-train.newspaper.png', 150, 60, 'contain', {'align': 'bottom', 'scale': 2}),
+    'ticket': ('last-train/closeups/last-train.ticket.closeup.png', 150, 80, 'contain', {'scale': 2}),
+    'departure-board': ('last-train/last-train.departure-board.png', 420, 200, 'contain', {'scale': 2}),
+    'carriage-plan': ('last-train/last-train.carriage-plan.png', 120, 150, 'contain', {'scale': 2}),
+    'carriage.props': ('last-train/last-train.carriage.props.png', 420, 300, 'contain', {'align': 'bottom'}),
+    'ledger': ('last-train/closeups/last-train.ledger.closeup.png', 170, 110, 'contain', {'scale': 2}),
+    'pocket-watch': ('last-train/closeups/last-train.pocket-watch.closeup.png', 160, 90, 'contain', {'scale': 2, 'face': {'cx': 72.8, 'cy': 53.3, 'd': 44}}),
+    'seat-card': ('last-train/last-train.seat-card.png', 110, 70, 'contain', {'scale': 2}),
+    'brake': ('last-train/last-train.brake.png', 140, 337, 'contain', {'scale': 2}),
+    'lever-frame': ('last-train/last-train.lever-frame.png', 900, 360, 'contain', {'align': 'bottom'}),
+    'lamp-room-door': ('last-train/last-train.lamp-room-door.png', 240, 440, 'contain', {'align': 'bottom'}),
+}
+
+
+def trim(im: Image.Image) -> Image.Image:
+    """Crop to the visible pixels."""
+    if im.mode != 'RGBA':
+        return im
+    box = im.getchannel('A').point(lambda v: 255 if v > 12 else 0).getbbox()
+    return im.crop(box) if box else im
+
+
+def cover(im, w, h):
+    s = max(w / im.width, h / im.height)
+    im = im.resize((round(im.width * s), round(im.height * s)), Image.LANCZOS)
+    x, y = (im.width - w) // 2, (im.height - h) // 2
+    return im.crop((x, y, x + w, y + h))
+
+
+def contain(im, w, h, align='center', stretch=False):
+    im = trim(im.convert('RGBA'))
+    if stretch:
+        # Tall or long thin things (a pole, a train) fill their box exactly.
+        return im.resize((w, h), Image.LANCZOS), (0, 0, w, h)
+    s = min(w / im.width, h / im.height)
+    nw, nh = max(1, round(im.width * s)), max(1, round(im.height * s))
+    im = im.resize((nw, nh), Image.LANCZOS)
+    canvas = Image.new('RGBA', (w, h), (0, 0, 0, 0))
+    x = (w - nw) // 2
+    y = h - nh if align == 'bottom' else (h - nh) // 2
+    canvas.alpha_composite(im, (x, y))
+    return canvas, (x, y, nw, nh)
+
+
+def fade_top(im, frac):
+    im = im.convert('RGBA')
+    a = np.array(im)
+    n = int(im.height * frac)
+    ramp = np.clip(np.linspace(0, 1, n) ** 1.6, 0, 1)
+    a[:n, :, 3] = (a[:n, :, 3] * ramp[:, None]).astype(np.uint8)
+    return Image.fromarray(a)
+
+
+def tile(im, w, h):
+    im = im.convert('RGB')
+    s = h / im.height
+    t = im.resize((round(im.width * s), h), Image.LANCZOS)
+    out = Image.new('RGB', (w, h))
+    x, i = 0, 0
+    while x < w:
+        out.paste(t if i % 2 == 0 else t.transpose(Image.FLIP_LEFT_RIGHT), (x, 0))
+        x += t.width
+        i += 1
+    return out
+
+
+def disc(im, w, h, at):
+    """Cut the round clock out of an opaque square painting. `at` = (cx, cy, r) as fractions of the width, measured by eye."""
+    im = im.convert('RGBA')
+    cx, cy, r = at[0] * im.width, at[1] * im.width, at[2] * im.width
+    mask = Image.new('L', im.size, 0)
+    ImageDraw.Draw(mask).ellipse((cx - r, cy - r, cx + r, cy + r), fill=255)
+    im.putalpha(mask.filter(ImageFilter.GaussianBlur(1.5)))
+    im = im.crop((int(cx - r), int(cy - r), int(cx + r), int(cy + r)))
+    return im.resize((w, h), Image.LANCZOS)
+
+
+def face_of(im, box_w):
+    """Centre and diameter of the light clock face, in box px."""
+    a = np.array(im.convert('RGBA')).astype(float)
+    lum = a[..., :3].mean(axis=2)
+    light = (lum > 140) & (a[..., 3] > 200)
+    lab, n = ndimage.label(ndimage.binary_closing(light, iterations=3))
+    if n == 0:
+        return None
+    sizes = ndimage.sum(light, lab, range(n + 1))
+    k = int(np.argmax(sizes[1:]) + 1)
+    ys, xs = np.where(lab == k)
+    scale = box_w / im.width
+    d = ((xs.max() - xs.min()) + (ys.max() - ys.min())) / 2
+    return {'cx': round(float((xs.min() + xs.max()) / 2 * scale), 1), 'cy': round(float((ys.min() + ys.max()) / 2 * scale), 1), 'd': round(float(d * scale), 1)}
+
+
+def save(im, path_stem):
+    has_alpha = im.mode == 'RGBA' and np.array(im.getchannel('A')).min() < 250
+    if has_alpha:
+        p = path_stem.parent / (path_stem.name + '.png')
+        im.save(p, optimize=True)
+    else:
+        p = path_stem.parent / (path_stem.name + '.jpg')
+        im.convert('RGB').save(p, quality=86, optimize=True, progressive=False)
+    return p
+
+
+def extras(src: Path):
+    """Clue close-ups (shown on evidence cards) and speech-bubble portraits."""
+    closeups = {
+        'platform-clock': 'clock', 'company-notice': 'notice', 'waiting-room-clock': 'waiting-room-clock',
+        'torn-ticket': 'ticket', 'departure-board': 'departure-board', 'carriage-plan': 'carriage-plan',
+        'pocket-watch': 'pocket-watch', 'conductors-ledger': 'ledger', 'seat-eight-card': 'seat-card',
+        'wired-brake': 'brake', 'blackwater-lever': 'lever-frame', 'lamp-room-door': 'lamp-room-door',
+    }
+    d = OUT / 'closeups'
+    d.mkdir(exist_ok=True)
+    for clue, name in closeups.items():
+        im = trim(Image.open(src / f'last-train/closeups/last-train.{name}.closeup.png').convert('RGBA'))
+        im.thumbnail((560, 400), Image.LANCZOS)
+        im.save(d / f'{clue}.png', optimize=True)
+        print(f'closeup {clue:20s} {im.size}')
+    # Portraits: the head and shoulders of each expression, square.
+    portraits = {
+        'voss': ('last-train/characters/last-train.voss.neutral.png', 'last-train/characters/last-train.voss.revealing.png'),
+        'driver': ('last-train/characters/last-train.driver.neutral.png', 'last-train/characters/last-train.driver.uneasy.png'),
+        'signalman': ('last-train/characters/last-train.signalman.nervous.png', 'last-train/characters/last-train.signalman.worried.png'),
+        'caldwell': ('last-train/characters/last-train.caldwell.neutral.png', 'last-train/characters/last-train.caldwell.shaken.png'),
+        'detective': ('shared/shared.grace.neutral.png', 'shared/shared.grace.thinking.png'),
+    }
+    d = OUT / 'portraits'
+    d.mkdir(exist_ok=True)
+    for who, files in portraits.items():
+        for mood, rel in zip(('calm', 'pressed'), files):
+            im = trim(Image.open(src / rel).convert('RGBA'))
+            a = np.array(im.getchannel('A')) > 40
+            top = int(np.argmax(a.any(axis=1)))
+            # Head: the opaque columns in the first 12% of the figure's height.
+            band = a[top:top + max(8, int(im.height * 0.12))]
+            cols = np.where(band.any(axis=0))[0]
+            cx = int((cols.min() + cols.max()) / 2)
+            # Grace's sources are half-length portraits; the others are full figures.
+            side = int(im.height * (0.5 if who == 'detective' else 0.27))
+            x0 = max(0, min(im.width - side, cx - side // 2)) if side <= im.width else cx - side // 2
+            box = (x0, max(0, top - int(side * 0.06)), x0 + side, max(0, top - int(side * 0.06)) + side)
+            sq = Image.new('RGBA', (side, side), (0, 0, 0, 0))
+            sq.alpha_composite(im.crop(box))
+            sq = sq.resize((264, 264), Image.LANCZOS)
+            sq.save(d / f'{who}.{mood}.png', optimize=True)
+            print(f'portrait {who}.{mood}')
+
+
+def main():
+    src = Path(sys.argv[1])
+    if '--extras' in sys.argv:
+        extras(src)
+        return
+    only = None
+    if '--only' in sys.argv:
+        only = set(sys.argv[sys.argv.index('--only') + 1].split(','))
+    OUT.mkdir(parents=True, exist_ok=True)
+    report_file = OUT / 'art-report.json'
+    report = json.loads(report_file.read_text()) if report_file.exists() else {}
+    for key, (rel, w, h, mode, o) in SPEC.items():
+        if only and key not in only:
+            continue
+        im = Image.open(src / rel)
+        k = o.get('scale', 1)
+        W, H = w * k, h * k
+        if mode == 'cover':
+            out = cover(im.convert('RGB'), W, H)
+            if o.get('fade'):
+                out = fade_top(out, o['fade'])
+        elif mode == 'band':
+            im = im.convert('RGB')
+            band = im.crop((0, int(im.height * o['from']), im.width, im.height))
+            out = fade_top(tile(band, W, H), o['fade'])
+        elif mode == 'tile':
+            out = tile(im, W, H)
+        elif mode == 'disc':
+            out = disc(im, W, H, o['at'])
+        else:
+            out, _ = contain(im, W, H, o.get('align', 'center'), o.get('stretch', False))
+        for ext in ('.png', '.jpg'):
+            (OUT / (key + ext)).unlink(missing_ok=True)
+        p = save(out, OUT / key)
+        entry = {'file': p.name, 'px': list(out.size), 'kb': round(p.stat().st_size / 1024)}
+        if o.get('face'):
+            # Where the procedural hands go (box px). Measured by eye on the
+            # painted dials; face_of() is the automatic first guess.
+            entry['face'] = o['face'] if isinstance(o['face'], dict) else face_of(out, w)
+        report[key] = entry
+        print(f'{key:22s} {p.name:28s} {out.size[0]}x{out.size[1]} {entry["kb"]:>5} KB {entry.get("face", "")}')
+    report_file.write_text(json.dumps(report, indent=2) + '\n')
+
+
+if __name__ == '__main__':
+    main()