Julian Bilcke
commited on
Commit
·
42642e4
1
Parent(s):
4c069c7
add the beta mode
Browse files
src/app/engine/render.ts
CHANGED
|
@@ -84,6 +84,8 @@ export async function newRender({
|
|
| 84 |
|
| 85 |
const placeholder = "<USE YOUR OWN TOKEN>"
|
| 86 |
|
|
|
|
|
|
|
| 87 |
// console.log("settings:", JSON.stringify(settings, null, 2))
|
| 88 |
|
| 89 |
if (
|
|
@@ -315,6 +317,7 @@ export async function newRender({
|
|
| 315 |
inputs: Buffer.from(blob).toString('base64'),
|
| 316 |
parameters: {
|
| 317 |
prompt: positivePrompt,
|
|
|
|
| 318 |
num_inference_steps: nbInferenceSteps,
|
| 319 |
guidance_scale: guidanceScale,
|
| 320 |
width,
|
|
@@ -369,7 +372,7 @@ export async function newRender({
|
|
| 369 |
},
|
| 370 |
body: JSON.stringify({
|
| 371 |
prompt,
|
| 372 |
-
negativePrompt
|
| 373 |
|
| 374 |
// for a future version of the comic factory
|
| 375 |
identityImage: "",
|
|
|
|
| 84 |
|
| 85 |
const placeholder = "<USE YOUR OWN TOKEN>"
|
| 86 |
|
| 87 |
+
const negativePrompt = "speech bubble, caption, subtitle"
|
| 88 |
+
|
| 89 |
// console.log("settings:", JSON.stringify(settings, null, 2))
|
| 90 |
|
| 91 |
if (
|
|
|
|
| 317 |
inputs: Buffer.from(blob).toString('base64'),
|
| 318 |
parameters: {
|
| 319 |
prompt: positivePrompt,
|
| 320 |
+
negative_prompt: negativePrompt,
|
| 321 |
num_inference_steps: nbInferenceSteps,
|
| 322 |
guidance_scale: guidanceScale,
|
| 323 |
width,
|
|
|
|
| 372 |
},
|
| 373 |
body: JSON.stringify({
|
| 374 |
prompt,
|
| 375 |
+
negativePrompt,
|
| 376 |
|
| 377 |
// for a future version of the comic factory
|
| 378 |
identityImage: "",
|
src/app/interface/panel/index.tsx
CHANGED
|
@@ -60,6 +60,8 @@ export function Panel({
|
|
| 60 |
|
| 61 |
const setPanelPrompt = useStore(s => s.setPanelPrompt)
|
| 62 |
|
|
|
|
|
|
|
| 63 |
const speeches = useStore(s => s.speeches)
|
| 64 |
const speech = speeches[panelIndex] || ""
|
| 65 |
const setPanelSpeech = useStore(s => s.setPanelSpeech)
|
|
@@ -96,17 +98,16 @@ export function Panel({
|
|
| 96 |
|
| 97 |
let delay = enableRateLimiter ? (1000 + (500 * panelIndex)) : 1000
|
| 98 |
|
| 99 |
-
|
| 100 |
-
|
| 101 |
const addSpeechBubble = async () => {
|
| 102 |
if (!renderedRef.current) { return }
|
| 103 |
|
| 104 |
// story generation failed
|
| 105 |
if (speech.trim() === "...") { return }
|
| 106 |
|
| 107 |
-
if (!
|
| 108 |
|
| 109 |
-
console.log('Generating speech
|
| 110 |
try {
|
| 111 |
const result = await injectSpeechBubbleInTheBackground({
|
| 112 |
inputImageInBase64: renderedRef.current.assetUrl,
|
|
|
|
| 60 |
|
| 61 |
const setPanelPrompt = useStore(s => s.setPanelPrompt)
|
| 62 |
|
| 63 |
+
const showSpeeches = useStore(s => s.showSpeeches)
|
| 64 |
+
|
| 65 |
const speeches = useStore(s => s.speeches)
|
| 66 |
const speech = speeches[panelIndex] || ""
|
| 67 |
const setPanelSpeech = useStore(s => s.setPanelSpeech)
|
|
|
|
| 98 |
|
| 99 |
let delay = enableRateLimiter ? (1000 + (500 * panelIndex)) : 1000
|
| 100 |
|
| 101 |
+
|
|
|
|
| 102 |
const addSpeechBubble = async () => {
|
| 103 |
if (!renderedRef.current) { return }
|
| 104 |
|
| 105 |
// story generation failed
|
| 106 |
if (speech.trim() === "...") { return }
|
| 107 |
|
| 108 |
+
if (!showSpeeches) { return }
|
| 109 |
|
| 110 |
+
console.log('Generating speech bubbles (this is experimental!)')
|
| 111 |
try {
|
| 112 |
const result = await injectSpeechBubbleInTheBackground({
|
| 113 |
inputImageInBase64: renderedRef.current.assetUrl,
|
src/app/interface/top-menu/index.tsx
CHANGED
|
@@ -179,8 +179,8 @@ export function TopMenu() {
|
|
| 179 |
onCheckedChange={setShowSpeeches}
|
| 180 |
/>
|
| 181 |
<Label className="text-gray-200 dark:text-gray-200">
|
| 182 |
-
<span className="hidden md:inline">
|
| 183 |
-
<span className="inline md:hidden">
|
| 184 |
</Label>
|
| 185 |
</div>
|
| 186 |
{/*
|
|
|
|
| 179 |
onCheckedChange={setShowSpeeches}
|
| 180 |
/>
|
| 181 |
<Label className="text-gray-200 dark:text-gray-200">
|
| 182 |
+
<span className="hidden md:inline">Beta</span>
|
| 183 |
+
<span className="inline md:hidden">Beta</span>
|
| 184 |
</Label>
|
| 185 |
</div>
|
| 186 |
{/*
|
src/lib/bubble/injectSpeechBubbleInTheBackground.ts
CHANGED
|
@@ -1,4 +1,5 @@
|
|
| 1 |
import { ImageSegmenter, FilesetResolver } from "@mediapipe/tasks-vision"
|
|
|
|
| 2 |
|
| 3 |
interface BoundingBox {
|
| 4 |
top: number;
|
|
@@ -25,7 +26,7 @@ export async function injectSpeechBubbleInTheBackground(params: {
|
|
| 25 |
text,
|
| 26 |
shape = "oval",
|
| 27 |
line = "handdrawn",
|
| 28 |
-
font =
|
| 29 |
debug = false,
|
| 30 |
} = params;
|
| 31 |
|
|
@@ -116,8 +117,11 @@ function analyzeSegmentationMask(mask: Uint8Array, width: number, height: number
|
|
| 116 |
}
|
| 117 |
|
| 118 |
function splitTextIntoBubbles(text: string): string[] {
|
| 119 |
-
|
| 120 |
-
|
|
|
|
|
|
|
|
|
|
| 121 |
}
|
| 122 |
|
| 123 |
function calculateBubbleLocations(
|
|
@@ -215,7 +219,7 @@ function drawSpeechBubble(
|
|
| 215 |
) {
|
| 216 |
const bubbleWidth = Math.min(300, imageWidth * 0.4);
|
| 217 |
const bubbleHeight = Math.min(150, imageHeight * 0.3);
|
| 218 |
-
const padding =
|
| 219 |
|
| 220 |
const fontSize = 20;
|
| 221 |
ctx.font = `${fontSize}px ${font}`;
|
|
@@ -232,7 +236,9 @@ function drawSpeechBubble(
|
|
| 232 |
};
|
| 233 |
|
| 234 |
ctx.fillStyle = 'white';
|
| 235 |
-
|
|
|
|
|
|
|
| 236 |
ctx.lineWidth = 2;
|
| 237 |
|
| 238 |
let tailTarget = null;
|
|
@@ -353,34 +359,60 @@ function drawTail(
|
|
| 353 |
tailTarget: { x: number, y: number },
|
| 354 |
shape: string
|
| 355 |
) {
|
| 356 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 357 |
const tailHeight = 30;
|
| 358 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 359 |
ctx.beginPath();
|
| 360 |
-
ctx.moveTo(
|
| 361 |
-
|
| 362 |
const controlPoint1 = {
|
| 363 |
-
x:
|
| 364 |
-
y:
|
| 365 |
};
|
| 366 |
-
|
| 367 |
const controlPoint2 = {
|
| 368 |
-
x:
|
| 369 |
-
y:
|
| 370 |
};
|
| 371 |
-
|
| 372 |
ctx.bezierCurveTo(
|
| 373 |
controlPoint1.x, controlPoint1.y,
|
| 374 |
controlPoint2.x, controlPoint2.y,
|
| 375 |
-
|
| 376 |
);
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 377 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 378 |
ctx.bezierCurveTo(
|
| 379 |
-
|
| 380 |
-
|
| 381 |
-
|
| 382 |
);
|
| 383 |
-
|
| 384 |
ctx.closePath();
|
| 385 |
ctx.fill();
|
| 386 |
ctx.stroke();
|
|
|
|
| 1 |
import { ImageSegmenter, FilesetResolver } from "@mediapipe/tasks-vision"
|
| 2 |
+
import { actionman } from "../fonts";
|
| 3 |
|
| 4 |
interface BoundingBox {
|
| 5 |
top: number;
|
|
|
|
| 26 |
text,
|
| 27 |
shape = "oval",
|
| 28 |
line = "handdrawn",
|
| 29 |
+
font = actionman.style.fontFamily,
|
| 30 |
debug = false,
|
| 31 |
} = params;
|
| 32 |
|
|
|
|
| 117 |
}
|
| 118 |
|
| 119 |
function splitTextIntoBubbles(text: string): string[] {
|
| 120 |
+
// Define a regular expression pattern
|
| 121 |
+
const pattern = /(?:[A-Z][a-z]*\.\s*)*(?:[^.!?\s]+[^.!?]*[.!?]+)|\S+/g;
|
| 122 |
+
|
| 123 |
+
const matches = text.match(pattern) || [text];
|
| 124 |
+
return matches.map(sentence => sentence.trim());
|
| 125 |
}
|
| 126 |
|
| 127 |
function calculateBubbleLocations(
|
|
|
|
| 219 |
) {
|
| 220 |
const bubbleWidth = Math.min(300, imageWidth * 0.4);
|
| 221 |
const bubbleHeight = Math.min(150, imageHeight * 0.3);
|
| 222 |
+
const padding = 24;
|
| 223 |
|
| 224 |
const fontSize = 20;
|
| 225 |
ctx.font = `${fontSize}px ${font}`;
|
|
|
|
| 236 |
};
|
| 237 |
|
| 238 |
ctx.fillStyle = 'white';
|
| 239 |
+
|
| 240 |
+
// let's disable the border for now
|
| 241 |
+
ctx.strokeStyle = 'white'; // 'black';
|
| 242 |
ctx.lineWidth = 2;
|
| 243 |
|
| 244 |
let tailTarget = null;
|
|
|
|
| 359 |
tailTarget: { x: number, y: number },
|
| 360 |
shape: string
|
| 361 |
) {
|
| 362 |
+
// Calculate new maximum length for tail
|
| 363 |
+
const bubbleCenterX = bubbleLocation.x;
|
| 364 |
+
const bubbleCenterY = bubbleLocation.y;
|
| 365 |
+
const maxTailLength = Math.max(bubbleHeight, bubbleWidth) * 0.6;
|
| 366 |
+
|
| 367 |
+
const tailBaseWidth = 30; // 50% larger than before
|
| 368 |
const tailHeight = 30;
|
| 369 |
+
|
| 370 |
+
// Calculate the length from bubble center to tail target
|
| 371 |
+
const deltaX = tailTarget.x - bubbleCenterX;
|
| 372 |
+
const deltaY = tailTarget.y - bubbleCenterY;
|
| 373 |
+
const distance = Math.sqrt(deltaX * deltaX + deltaY * deltaY);
|
| 374 |
+
|
| 375 |
+
// Normalize the length if it exceeds the max tail length
|
| 376 |
+
const limitedDistance = Math.min(distance, maxTailLength);
|
| 377 |
+
const tailEndX = bubbleCenterX + (deltaX / distance) * limitedDistance;
|
| 378 |
+
const tailEndY = bubbleCenterY + (deltaY / distance) * limitedDistance;
|
| 379 |
+
|
| 380 |
ctx.beginPath();
|
| 381 |
+
ctx.moveTo(bubbleCenterX, bubbleCenterY);
|
| 382 |
+
|
| 383 |
const controlPoint1 = {
|
| 384 |
+
x: bubbleCenterX + deltaX / 3,
|
| 385 |
+
y: bubbleCenterY + deltaY / 3
|
| 386 |
};
|
| 387 |
+
|
| 388 |
const controlPoint2 = {
|
| 389 |
+
x: bubbleCenterX + (deltaX * 2) / 3,
|
| 390 |
+
y: bubbleCenterY + (deltaY * 2) / 3
|
| 391 |
};
|
| 392 |
+
|
| 393 |
ctx.bezierCurveTo(
|
| 394 |
controlPoint1.x, controlPoint1.y,
|
| 395 |
controlPoint2.x, controlPoint2.y,
|
| 396 |
+
tailEndX, tailEndY
|
| 397 |
);
|
| 398 |
+
|
| 399 |
+
// Mirror to create the width at base
|
| 400 |
+
const mirroredControlPoint1 = {
|
| 401 |
+
x: controlPoint1.x + tailBaseWidth / 3,
|
| 402 |
+
y: controlPoint1.y
|
| 403 |
+
};
|
| 404 |
|
| 405 |
+
const mirroredControlPoint2 = {
|
| 406 |
+
x: controlPoint2.x + (tailBaseWidth * 2) / 3,
|
| 407 |
+
y: controlPoint2.y
|
| 408 |
+
};
|
| 409 |
+
|
| 410 |
ctx.bezierCurveTo(
|
| 411 |
+
mirroredControlPoint2.x, mirroredControlPoint2.y,
|
| 412 |
+
mirroredControlPoint1.x, mirroredControlPoint1.y,
|
| 413 |
+
bubbleCenterX + tailBaseWidth, bubbleCenterY
|
| 414 |
);
|
| 415 |
+
|
| 416 |
ctx.closePath();
|
| 417 |
ctx.fill();
|
| 418 |
ctx.stroke();
|