Animate Lumi wing strokes and narrate the promo through OpenRouter

This commit is contained in:
waozi committed 2026-10-09 00:01:11 -03:00
1 parent 284282fc10
commit 883f0eff65
15 files changed
+304 -54

No files matched your search

+4 -1
View File
@@ -1173,7 +1173,7 @@ social-post-dry-run:
screenshot: $(TARGET)
./scripts/generate-screenshots.sh "$(TARGET)"
.PHONY: screenshot-verify screenshot-review screenshot-export screenshot-source-check play-preflight lumi-promo
.PHONY: screenshot-verify screenshot-review screenshot-export screenshot-source-check play-preflight lumi-promo lumi-narration
screenshot-source-check: $(ZI2C_BIN) | build-laws
@ZI_CHECK_DEFINES="$(ZI_NATIVE_DEFINES)" python3 scripts/check-zi-sources.py $(ZI2ZIR_BIN) $(KRYON_DIR)/src/ui $(ZIRAN_DIR)/std $(KSS_DIR)/src
@@ -1192,6 +1192,9 @@ play-preflight:
lumi-promo:
python3 scripts/render-lumi-promo.py
lumi-narration:
python3 scripts/generate-lumi-narration.py
test-tui-screenshot test-termi-screenshot:
@$(MAKE) --no-print-directory KRYON_BACKEND=termi test-termi-screenshot-direct
+7 -2
View File
@@ -115,9 +115,14 @@ and the existing release workflow, which builds and publishes the signed
Google Play AAB and owns release tags. A local build or commit is not a Play
deployment.
`make lumi-promo` renders a 30-second portrait review MP4 from the exact
`make lumi-promo` renders a narrated portrait review MP4 from the exact
reviewed phone captures and the generated artwork in `design/lumi-promo/`.
Lumi flies between screens with light trails and original synthesized bells.
Lumi uses eight generated wing-flapping frames, flies to each screen and settles
while a narrator explains it. The audio contains speech only, without music or
sound effects. Scene timings expand to fit the complete spoken lines.
`make lumi-narration` generates the stored narrator clips through OpenRouter
using a locally configured `OPENROUTER_API_KEY`; unchanged clips are reused.
The voice, script and provider options are in `design/lumi-promo/narration.json`.
The chat example runs real local theme actions; the surrounding forest and
glow are marketing effects. Inspect `build/lumi-promo/storyboard.jpg` and
`build/lumi-promo/inner-breeze-lumi-review.mp4` before using it publicly.
+8
View File
@@ -9,5 +9,13 @@
"transparent": false,
"prompt": "Use case: ads-marketing. Asset type: portrait background plate for an Inner Breeze wellness app promo guided by Lumi, a glowing golden firefly. Create an enchanting cinematic forest glade at twilight in a premium soft storybook painted 3D style with delicate watercolor texture. Deep sage, teal, moss and midnight green foliage framing only the extreme bottom and side edges, many tiny warm gold firefly lights at different depths, soft volumetric moonlight and gentle golden light trails like breath rising toward the center. Calm spacious central area with an uncluttered deep teal-to-sage atmospheric gradient, where readable app screens will be composited later. Magical intimate and inviting, lush organic detail, exquisite soft glowing bokeh, subtle painterly paper texture. Portrait 9:16 composition. No characters, no letters or typography, no phones, no UI, no logos or watermark."
},
"lumi-flight-cycle.png": {
"reference": "lumi-flight.png",
"transparent": true,
"columns": 4,
"rows": 2,
"frames": 8,
"prompt": "Use case: ads-marketing. Asset type: transparent 8-frame wing-flapping sprite sheet for an animated video. Input image 1 is the Lumi identity reference; input image 2 is the existing animation sheet to fix. Recreate image 2 as an evenly spaced 4-column by 2-row grid of exactly 8 full-body frames, same warm olive green firefly Lumi and same soft painted 3D texture. Critical correction: shrink EVERY character so its complete antenna glow, wings and feet occupy ONLY the central 60 percent of its cell, leaving an empty transparent margin of at least 18 percent along EVERY side. No character part or glow can touch a cell edge. Keep precisely the same head center and body size in all 8 cells. Body stable; wings do real full up/down strokes. Sequential wing positions left-to-right then second row: frame 1 wings up, frame 2 middle downward, frame 3 wings fully down, frame 4 middle upward, frame 5 wings up, frame 6 middle downward, frame 7 wings fully down, frame 8 middle upward. Clearly distinct full strokes, not duplicate wings. Both delicate wings attached at the back, correct anatomy. No whole-body rotation or sliding. Preserve face and hands identically so playback doesn't flicker. Tiny feet gently flex with the strokes. Big friendly brown eyes and inviting smile. Golden glows localized close to antenna bulbs and belly, no large halo clouds. True transparent alpha everywhere outside the character. No scenery, no grid lines, no text or labels. Wide canvas 2:1 aspect ratio, high resolution. Character always fully contained within its allotted cell with very generous empty alpha padding."
},
"status": "Review artwork for a promo; does not change the in-app mascot or claim new customization capabilities."
}
Binary file not shown.

After

Width:  |  Height:  |  Size: 1.2 MiB

+17
View File
@@ -0,0 +1,17 @@
{
"model": "microsoft/mai-voice-2.1",
"voice": "en-US-Harper:MAI-Voice-2.1",
"response_format": "mp3",
"speed": 1.0,
"provider": {"options": {"azure": {"style": "narrator", "styledegree": 1.0}}},
"clips": [
"Meet Lumi. Your little guide to Inner Breeze.",
"A space for your apps, and your daily rhythm.",
"Take a breath. Find your calm. Make time to move.",
"Build little habits that make every day brighter.",
"Keep your plans in Lists, and your thoughts in Diary.",
"Make it yours. Ask Lumi for a forest theme, then switch to dark mode.",
"Inner Breeze. A little space, made for you."
],
"documentation": "https://openrouter.ai/docs/guides/overview/multimodal/tts"
}
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
+59
View File
@@ -0,0 +1,59 @@
{
"config_sha256": "a3f0fbaea5ab604d86f9117572956e0597f18d314660c16c836e38ed431e1c83",
"provider": "OpenRouter",
"model": "microsoft/mai-voice-2.1",
"voice": "en-US-Harper:MAI-Voice-2.1",
"clips": [
{
"file": "clip-00.mp3",
"text": "Meet Lumi. Your little guide to Inner Breeze.",
"seconds": 3.648,
"sha256": "31777f316cb9f786ea69a051be06fd82ad4a8e7508a1f739a9b6faa8a3dd4734",
"generation_id": "gen-tts-1791514279-Tg7LWgrcfvxt2OIWjuDj"
},
{
"file": "clip-01.mp3",
"text": "A space for your apps, and your daily rhythm.",
"seconds": 2.952,
"sha256": "766bdb7e9ca8c44b6690170ab20ac61ddfcca4934947d4f369d150ea2a6e7224",
"generation_id": "gen-tts-1791514280-LacvzXxRZsGdXgjW1xmY"
},
{
"file": "clip-02.mp3",
"text": "Take a breath. Find your calm. Make time to move.",
"seconds": 4.752,
"sha256": "f2ed561fa0b3c77775b4cc3b3a4fb19fa1ea6cb23cc0bbc4c7216305edd12f4f",
"generation_id": "gen-tts-1791514282-FVKXyQSwJ0oGJecZ9MPL"
},
{
"file": "clip-03.mp3",
"text": "Build little habits that make every day brighter.",
"seconds": 2.784,
"sha256": "3685d1e6ec34bd1c2bbd98db10b0fd3e0cc3313bb14dfd1ccae596289eea849b",
"generation_id": "gen-tts-1791514283-tIeLAmKNQfvB5vqNpjLR"
},
{
"file": "clip-04.mp3",
"text": "Keep your plans in Lists, and your thoughts in Diary.",
"seconds": 3.48,
"sha256": "34db8fdb37793e22425867fc5123d0abdac4cf8fa515f373b1553e7c4d31bef0",
"generation_id": "gen-tts-1791514286-Be2DMWRxLmqQHWnLy2WK"
},
{
"file": "clip-05.mp3",
"text": "Make it yours. Ask Lumi for a forest theme, then switch to dark mode.",
"seconds": 5.088,
"sha256": "3a9816c6484ee089bdfd36bc15b8efd8d925b79ad4477c8b3680a3918706e4c5",
"generation_id": "gen-tts-1791514288-kfGqTRkfR6gYKGQzSzTH"
},
{
"file": "clip-06.mp3",
"text": "Inner Breeze. A little space, made for you.",
"seconds": 4.056,
"sha256": "55e9a0c3b19a8c5b6f9eaf23bb81b1b3cc67a0a6b864f3aa57ad3103ef0c3eaa",
"generation_id": "gen-tts-1791514290-ttQk1lQ5H3modKGaq95J"
}
],
"music": false,
"documentation": "https://openrouter.ai/docs/guides/overview/multimodal/tts"
}
+99
View File
@@ -0,0 +1,99 @@
#!/usr/bin/env python3
"""Generate the promo's spoken narration through OpenRouter, without music."""
import argparse
import hashlib
import json
import os
from pathlib import Path
import subprocess
import tempfile
import urllib.error
import urllib.request
ROOT = Path(__file__).resolve().parents[1]
ASSETS = ROOT / "design/lumi-promo"
ENDPOINT = "https://openrouter.ai/api/v1/audio/speech"
def digest(path):
return hashlib.sha256(path.read_bytes()).hexdigest()
def duration(path):
return float(subprocess.check_output([
"ffprobe", "-v", "error", "-show_entries", "format=duration",
"-of", "default=noprint_wrappers=1:nokey=1", str(path)], text=True))
def main():
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("--regenerate", action="store_true")
args = parser.parse_args()
config_path = ASSETS / "narration.json"
config = json.loads(config_path.read_text())
output = ASSETS / "narration"
output.mkdir(exist_ok=True)
receipt_path = output / "receipt.json"
if receipt_path.is_file() and not args.regenerate:
receipt = json.loads(receipt_path.read_text())
if receipt["config_sha256"] == digest(config_path) and all(
(output / row["file"]).is_file() and
digest(output / row["file"]) == row["sha256"]
for row in receipt["clips"]):
print("Using the existing verified OpenRouter narration.")
return
key = os.environ.get("OPENROUTER_API_KEY", "")
if not key:
raise SystemExit("Set OPENROUTER_API_KEY locally before generating narration.")
models_url = "https://openrouter.ai/api/v1/models?output_modalities=speech"
with urllib.request.urlopen(models_url, timeout=30) as response:
models = json.load(response)["data"]
model = next((item for item in models if item["id"] == config["model"]), None)
if model is None or config["voice"] not in model.get("supported_voices", []):
raise SystemExit("The configured narrator model or voice is not in OpenRouter's current catalog.")
rows = []
for index, text in enumerate(config["clips"]):
body = {name: config[name] for name in (
"model", "voice", "response_format", "speed", "provider")}
body["input"] = text
request = urllib.request.Request(ENDPOINT,
data=json.dumps(body).encode(), headers={
"Authorization": "Bearer " + key,
"Content-Type": "application/json",
"X-Title": "Inner Breeze Lumi promo narration"})
try:
with urllib.request.urlopen(request, timeout=90) as response:
mime = response.headers.get_content_type()
if not mime.startswith("audio/"):
raise ValueError("OpenRouter returned a non-audio response")
audio = response.read()
generation = response.headers.get("X-Generation-Id")
except urllib.error.HTTPError as error:
raise SystemExit(f"OpenRouter narration request failed (HTTP {error.code}); no audio was saved.") from None
if len(audio) < 1024:
raise ValueError("OpenRouter returned empty or truncated narration")
destination = output / f"clip-{index:02}.mp3"
with tempfile.NamedTemporaryFile(suffix=".mp3", dir=output, delete=False) as stream:
stream.write(audio)
temporary = Path(stream.name)
try:
seconds = duration(temporary)
if not 0.3 <= seconds <= 20:
raise ValueError("Narration clip duration is outside the expected range")
subprocess.run(["ffmpeg", "-v", "error", "-i", str(temporary),
"-f", "null", "-"], check=True, capture_output=True)
temporary.replace(destination)
finally:
temporary.unlink(missing_ok=True)
rows.append(dict(file=destination.name, text=text, seconds=seconds,
sha256=digest(destination), generation_id=generation))
print(f"Generated narrator line {index + 1}/{len(config['clips'])}: {seconds:.2f} seconds", flush=True)
receipt = dict(config_sha256=digest(config_path), provider="OpenRouter",
model=config["model"], voice=config["voice"], clips=rows,
music=False, documentation=config["documentation"])
receipt_path.write_text(json.dumps(receipt, indent=2) + "\n")
print("Saved all narrator clips and their generation record.")
if __name__ == "__main__":
main()
+110 -51
View File
@@ -1,6 +1,7 @@
#!/usr/bin/env python3
"""Render a captioned Lumi tour using the exact visually reviewed captures."""
import argparse
from collections import deque
import hashlib
import importlib.util
import json
@@ -11,7 +12,7 @@ import subprocess
import wave
import numpy as np
from PIL import Image, ImageDraw, ImageFilter, ImageFont
from PIL import Image, ImageDraw, ImageFilter, ImageFont, ImageOps
ROOT = Path(__file__).resolve().parents[1]
ASSETS = ROOT / "design/lumi-promo"
@@ -39,23 +40,82 @@ def ease(value):
return value * value * (3 - 2 * value)
def soundtrack(path, seconds):
"""An original quiet bell bed, synthesized here; no licensed music."""
def narrated_story(path):
"""Leave each spoken line enough time; never trim speech to fit a card."""
config_path = ASSETS / "narration.json"
narration = json.loads((ASSETS / "narration/receipt.json").read_text())
assert narration["config_sha256"] == hashlib.sha256(config_path.read_bytes()).hexdigest()
assert len(narration["clips"]) == len(STORY)
rate = 48000
time = np.arange(seconds * rate) / rate
sound = np.zeros_like(time)
for at, frequency in [(0, 261.63), (4, 329.63), (8, 392), (12, 523.25), (16, 392), (20, 329.63), (23, 523.25), (26, 261.63), (27, 392)]:
offset = time - at
active = offset >= 0
envelope = np.exp(-np.maximum(offset, 0) / 2.7) * np.minimum(np.maximum(offset, 0) * 12, 1)
sound += active * envelope * (np.sin(2 * math.pi * frequency * offset) + .22 * np.sin(2 * math.pi * frequency * 2.01 * offset)) * .075
sound *= np.minimum(time, 1) * np.minimum(np.maximum(seconds - time, 0), 1)
stereo = np.stack([sound, sound * .94], axis=1)
story = []
clips = []
start = 0.0
for beat, row in zip(STORY, narration["clips"]):
audio = ASSETS / "narration" / row["file"]
assert hashlib.sha256(audio.read_bytes()).hexdigest() == row["sha256"], "Narration changed; regenerate its receipt"
raw = subprocess.check_output([
"ffmpeg", "-v", "error", "-i", str(audio), "-af",
"loudnorm=I=-16:TP=-1.5:LRA=7", "-ar", str(rate), "-ac", "1",
"-f", "f32le", "pipe:1"])
samples = np.frombuffer(raw, dtype="<f4")
end = start + max(beat[1] - beat[0], len(samples) / rate + .65)
story.append((start, end, *beat[2:]))
clips.append((round((start + .25) * rate), samples))
start = end
sound = np.zeros(math.ceil(start * rate), dtype=np.float32)
for offset, samples in clips:
sound[offset:offset + len(samples)] = samples
with wave.open(str(path), "wb") as stream:
stream.setnchannels(2)
stream.setnchannels(1)
stream.setsampwidth(2)
stream.setframerate(rate)
stream.writeframes((stereo * 32767).astype("<i2").tobytes())
stream.writeframes((np.clip(sound, -1, 1) * 32767).astype("<i2").tobytes())
return story, start, narration
def animation_frames():
"""Slice the generated cycle and align eyes to avoid registration jitter."""
atlas = Image.open(ASSETS / "lumi-flight-cycle.png").convert("RGBA")
assert atlas.getchannel("A").getextrema() == (0, 255), "Lumi needs true transparency"
tiles = []
anchors = []
side = round(atlas.width / 4)
for index in range(8):
column, row = index % 4, index // 4
box = (round(column * atlas.width / 4), round(row * atlas.height / 2),
round((column + 1) * atlas.width / 4), round((row + 1) * atlas.height / 2))
tile = atlas.crop(box).resize((side, side), Image.Resampling.LANCZOS)
pixels = np.asarray(tile)
eyes = ((pixels[:, :, 0] < 65) & (pixels[:, :, 1] < 65) &
(pixels[:, :, 2] < 60) & (pixels[:, :, 3] > 220))
eyes[:int(side * .28)] = False
eyes[int(side * .66):] = False
eyes[:, :int(side * .38)] = False
eyes[:, int(side * .81):] = False
ys, xs = np.nonzero(eyes)
assert len(xs) > 100, "Animation frame has no stable eye anchor"
anchors.append((float(xs.mean()), float(ys.mean())))
tiles.append(tile)
target = np.mean(anchors, axis=0)
aligned = []
for tile, anchor in zip(tiles, anchors):
canvas = Image.new("RGBA", tile.size)
canvas.alpha_composite(tile, (round(target[0] - anchor[0]), round(target[1] - anchor[1])))
aligned.append(canvas)
return aligned
def flight_position(index, local, width, height):
"""Fly to a destination, settle there, then move when the scene changes."""
targets = [(.5, .48), (.85, .66), (.85, .54), (.85, .66),
(.85, .58), (.85, .65), (.5, .48)]
origin = (-.3, .58) if index == 0 else targets[index - 1]
target = targets[index]
progress = ease(local / 1.35)
x = origin[0] + (target[0] - origin[0]) * progress
y = origin[1] + (target[1] - origin[1]) * progress
y -= math.sin(progress * math.pi) * .045
return x * width, y * height, progress
def main():
@@ -74,12 +134,10 @@ def main():
width, height = args.width, args.height
scale = width / 1080
background = Image.open(ASSETS / "forest.png").convert("RGB")
from PIL import ImageOps
background = ImageOps.fit(background, (width, height)).convert("RGBA")
shade = Image.new("RGBA", (width, height), (3, 19, 20, 90))
background = Image.alpha_composite(background, shade)
lumi = Image.open(ASSETS / "lumi-flight.png").convert("RGBA")
assert lumi.getchannel("A").getextrema() == (0, 255), "Lumi needs true transparency"
sprites = animation_frames()
title_font = ImageFont.truetype(BOLD, int(54 * scale))
small_font = ImageFont.truetype(FONT, int(29 * scale))
brand_font = ImageFont.truetype(FONT, int(23 * scale))
@@ -94,17 +152,20 @@ def main():
cards[name] = picture
randomizer = random.Random(42)
motes = [(randomizer.random(), randomizer.random(), randomizer.uniform(.4, 1.4), randomizer.random() * math.tau) for _ in range(70)]
audio = OUTPUT / "lumi-bells.wav"
soundtrack(audio, 30)
audio = OUTPUT / "lumi-narration.wav"
story, seconds, narration = narrated_story(audio)
destination = OUTPUT / "inner-breeze-lumi-review.mp4"
command = ["ffmpeg", "-hide_banner", "-loglevel", "error", "-y", "-f", "rawvideo", "-pixel_format", "rgb24", "-video_size", f"{width}x{height}", "-framerate", str(args.fps), "-i", "pipe:0", "-i", str(audio), "-c:v", "libx264", "-preset", "fast", "-crf", "19", "-pix_fmt", "yuv420p", "-c:a", "aac", "-b:a", "160k", "-movflags", "+faststart", "-shortest", str(destination)]
staged = OUTPUT / "inner-breeze-lumi-rendering.mp4"
command = ["ffmpeg", "-hide_banner", "-loglevel", "error", "-y", "-f", "rawvideo", "-pixel_format", "rgb24", "-video_size", f"{width}x{height}", "-framerate", str(args.fps), "-i", "pipe:0", "-i", str(audio), "-c:v", "libx264", "-preset", "fast", "-crf", "19", "-pix_fmt", "yuv420p", "-c:a", "aac", "-b:a", "160k", "-movflags", "+faststart", "-shortest", str(staged)]
review_frames = []
trail = deque(maxlen=24)
with (OUTPUT / "render.log").open("w") as log:
encoder = subprocess.Popen(command, stdin=subprocess.PIPE, stderr=log)
try:
for frame in range(30 * args.fps):
for frame in range(math.ceil(seconds * args.fps)):
time = frame / args.fps
start, end, title, subtitle, scene = next(beat for beat in STORY if beat[0] <= time < beat[1])
index, beat = next((i, beat) for i, beat in enumerate(story) if beat[0] <= time < beat[1])
start, end, title, subtitle, scene = beat
local = time - start
canvas = background.copy()
draw = ImageDraw.Draw(canvas, "RGBA")
@@ -119,70 +180,67 @@ def main():
text_center(draw, subtitle, int(290 * scale), small_font, "#dfe9d8", width)
if scene:
card = cards[scene]
if start == 20:
card = Image.blend(cards["09-lumi-light"], cards[scene], ease((local - 2.5) / .8))
if index == 5:
card = Image.blend(cards["09-lumi-light"], cards[scene], ease((local / (end - start) - .43) / .14))
left = (width - card_width) // 2
top = int(410 * scale + 12 * scale * math.sin(time * .8))
top = int(410 * scale)
top += int((1 - ease(local / .7)) * 130 * scale)
shadow = Image.new("RGBA", canvas.size)
ImageDraw.Draw(shadow).rounded_rectangle((left - 12, top - 12, left + card_width + 12, top + card_height + 12), radius=int(48 * scale), fill=(193, 218, 150, 115))
canvas = Image.alpha_composite(canvas, shadow.filter(ImageFilter.GaussianBlur(int(22 * scale))))
canvas.alpha_composite(card, (left, top))
# An orbiting guide and a glowing trail reveal each screen.
progress = min(local / (end - start), 1)
center_x = width * (.79 - .13 * math.sin(progress * math.pi))
center_y = height * (.72 - .37 * progress)
character_size = int(260 * scale)
else:
center_x = width * .5 + width * .13 * math.sin(local * 1.05)
center_y = height * (.48 + .025 * math.sin(time * 2.2))
character_size = int((570 if start == 0 else 500) * scale)
center_x, center_y, progress = flight_position(index, local, width, height)
previous_size = 570 if index <= 1 else 285
target_size = 570 if index in (0, 6) else 285
character_size = int((previous_size + (target_size - previous_size) * progress) * scale)
draw = ImageDraw.Draw(canvas, "RGBA")
for tail in range(32, 0, -1):
px = center_x - tail * 5 * scale
py = center_y + math.sin(time * 4 - tail * .14) * 18 * scale + tail * scale
radius = (4 - tail / 12) * scale
draw.ellipse((px - radius, py - radius, px + radius, py + radius), fill=(255, 222, 122, int(120 * (1 - tail / 33))))
character = lumi.resize((character_size, character_size), Image.Resampling.LANCZOS)
character = character.rotate(5 * math.sin(time * 3), resample=Image.Resampling.BICUBIC, expand=True)
trail.append((center_x, center_y))
for age, (px, py) in enumerate(trail):
if math.hypot(px - center_x, py - center_y) > 8 * scale:
radius = (1 + age / len(trail) * 2) * scale
draw.ellipse((px - radius, py - radius, px + radius, py + radius), fill=(255, 222, 122, int(100 * age / len(trail))))
phase = (time * 12) % len(sprites)
pose = int(phase)
character = Image.blend(sprites[pose], sprites[(pose + 1) % len(sprites)], ease(phase - pose))
character = character.resize((character_size, character_size), Image.Resampling.LANCZOS)
glow = Image.new("RGBA", canvas.size)
glow_draw = ImageDraw.Draw(glow)
radius = character_size * .31
glow_draw.ellipse((center_x - radius, center_y - radius, center_x + radius, center_y + radius), fill=(255, 212, 105, 75))
canvas = Image.alpha_composite(canvas, glow.filter(ImageFilter.GaussianBlur(int(42 * scale))))
canvas.alpha_composite(character, (int(center_x - character.width / 2), int(center_y - character.height / 2)))
if start == 16 and local < 1.6:
if index == 4 and local < 1.6:
thumbnail = cards["06-lists"].resize((int(width * .33), int(height * .33)), Image.Resampling.LANCZOS)
canvas.alpha_composite(thumbnail, (int(width * .03), int(height * .5)))
if start == 20:
if index == 5:
draw = ImageDraw.Draw(canvas, "RGBA")
pill = "theme forest" if local < 3 else "/dark"
pill = "theme forest" if local / (end - start) < .5 else "/dark"
label_font = ImageFont.truetype(BOLD, int(32 * scale))
pill_width = int(draw.textlength(pill, font=label_font) + 70 * scale)
left = (width - pill_width) // 2
draw.rounded_rectangle((left, int(1670 * scale), left + pill_width, int(1760 * scale)), radius=int(35 * scale), fill=(231, 239, 210, 245))
text_center(draw, pill, int(1693 * scale), label_font, "#214332", width)
if start == 26:
if index == 6:
draw = ImageDraw.Draw(canvas, "RGBA")
text_center(draw, "Breathe · Grow · Make it yours", int(1390 * scale), small_font, "#fff0c2", width)
fade = min(ease(time / .5), ease((30 - time) / .8))
fade = min(ease(time / .5), ease((seconds - time) / .8))
if fade < 1:
canvas = Image.blend(Image.new("RGBA", canvas.size, "#071c1b"), canvas, fade)
if abs(local - 1.5) < .5 / args.fps:
path = OUTPUT / f"review-{start:02}.jpg"
path = OUTPUT / f"review-{index:02}.jpg"
canvas.convert("RGB").save(path, quality=93)
review_frames.append(path)
encoder.stdin.write(canvas.convert("RGB").tobytes())
if frame % (args.fps * 4) == 0:
print(f"Rendered {time:.0f}/30 seconds", flush=True)
print(f"Rendered {time:.0f}/{seconds:.1f} seconds", flush=True)
finally:
encoder.stdin.close()
assert encoder.wait() == 0, "video encoding failed; inspect render.log"
harness.verify(reviewed=True, current=False)
probe = json.loads(subprocess.check_output(["ffprobe", "-v", "error", "-show_streams", "-show_format", "-of", "json", str(destination)], text=True))
probe = json.loads(subprocess.check_output(["ffprobe", "-v", "error", "-show_streams", "-show_format", "-of", "json", str(staged)], text=True))
video = next(stream for stream in probe["streams"] if stream["codec_type"] == "video")
assert (video["width"], video["height"]) == (width, height) and video["codec_name"] == "h264"
assert abs(float(probe["format"]["duration"]) - 30) < .2
assert abs(float(probe["format"]["duration"]) - seconds) < .2
assert any(stream["codec_type"] == "audio" for stream in probe["streams"])
sheet = Image.new("RGB", (280 * len(review_frames), 500), "#09221d")
for index, path in enumerate(review_frames):
@@ -190,7 +248,8 @@ def main():
picture.thumbnail((280, 500))
sheet.paste(picture, (index * 280, 0))
sheet.save(OUTPUT / "storyboard.jpg", quality=95)
receipt = dict(capture_sha256=harness.digest(harness.OUTPUT / "capture.json"), artwork={name: harness.digest(ASSETS / name) for name in ("lumi-flight.png", "forest.png")}, video_sha256=harness.digest(destination), seconds=30, width=width, height=height, fps=args.fps, status="owner-review", soundtrack="original synthesized bells", story=STORY)
staged.replace(destination)
receipt = dict(capture_sha256=harness.digest(harness.OUTPUT / "capture.json"), artwork={name: harness.digest(ASSETS / name) for name in ("lumi-flight-cycle.png", "forest.png")}, video_sha256=harness.digest(destination), seconds=seconds, width=width, height=height, fps=args.fps, status="owner-review", soundtrack="OpenRouter narrator only; no music or sound effects", narration_sha256=harness.digest(ASSETS / "narration/receipt.json"), narrator=dict(provider=narration["provider"], model=narration["model"], voice=narration["voice"]), animation_frames=len(sprites), story=story)
harness.write_json(OUTPUT / "render-receipt.json", receipt)
audio.unlink()
print(f"Review video saved: {destination}")