Add reference images from previous page and story characters for consist

This commit is contained in:
Riccardo Giorato
2025-12-26 18:08:21 +01:00
parent 7fd5367577
commit 85f18f0ed1
6 changed files with 278 additions and 219 deletions
+21 -1
View File
@@ -6,6 +6,8 @@ import {
createPage,
getNextPageNumber,
getStoryWithPagesBySlug,
getLastPageImage,
getStoryCharacterImages,
} from "@/lib/db-actions";
import { freeTierRateLimit } from "@/lib/rate-limit";
import { uploadImageToS3 } from "@/lib/s3-upload";
@@ -84,6 +86,24 @@ export async function POST(request: NextRequest) {
const dimensions = FIXED_DIMENSIONS;
// Collect reference images: previous page + story characters + current characters
let referenceImages: string[] = [];
// Get previous page image for style consistency (unless it's page 1)
if (nextPageNumber > 1) {
const lastPageImage = await getLastPageImage(story.id);
if (lastPageImage) {
referenceImages.push(lastPageImage);
}
}
// Get story character images (up to 2 most recent)
const storyCharacterImages = await getStoryCharacterImages(story.id);
referenceImages.push(...storyCharacterImages.slice(-2)); // Take last 2
// Add current character images to references
referenceImages.push(...characterImages);
// Build the prompt with continuation context
const previousPages = pages.map(p => ({
prompt: p.prompt,
@@ -108,7 +128,7 @@ export async function POST(request: NextRequest) {
width: dimensions.width,
height: dimensions.height,
temperature: 0.1,
reference_images: characterImages.length > 0 ? characterImages : undefined,
reference_images: referenceImages.length > 0 ? referenceImages : undefined,
});
} catch (error) {
console.error("Together AI API error:", error);
+22 -3
View File
@@ -7,6 +7,8 @@ import {
createPage,
getNextPageNumber,
getStoryById,
getLastPageImage,
getStoryCharacterImages,
} from "@/lib/db-actions";
import { freeTierRateLimit } from "@/lib/rate-limit";
import { COMIC_STYLES } from "@/lib/constants";
@@ -89,9 +91,11 @@ export async function POST(request: NextRequest) {
let page;
let story;
let referenceImages: string[] = [];
if (storyId) {
const story = await getStoryById(storyId);
// Continuation: get previous page image and story character images
story = await getStoryById(storyId);
if (!story) {
return NextResponse.json({ error: "Story not found" }, { status: 404 });
}
@@ -103,7 +107,20 @@ export async function POST(request: NextRequest) {
prompt,
characterImageUrls: characterImages,
});
// Get previous page image for style consistency (unless it's page 1)
if (nextPageNumber > 1) {
const lastPageImage = await getLastPageImage(storyId);
if (lastPageImage) {
referenceImages.push(lastPageImage);
}
}
// Get story character images (up to 2 most recent)
const storyCharacterImages = await getStoryCharacterImages(storyId);
referenceImages.push(...storyCharacterImages.slice(-2)); // Take last 2
} else {
// New story: no previous page reference
story = await createStory({
title: prompt.length > 50 ? prompt.substring(0, 50) + "..." : prompt,
description: undefined,
@@ -119,6 +136,9 @@ export async function POST(request: NextRequest) {
});
}
// Add current character images to references
referenceImages.push(...characterImages);
const dimensions = FIXED_DIMENSIONS;
const fullPrompt = buildComicPrompt({
@@ -139,8 +159,7 @@ export async function POST(request: NextRequest) {
width: dimensions.width,
height: dimensions.height,
temperature: 0.1, // Lower temperature for more consistent face matching
reference_images:
characterImages.length > 0 ? characterImages : undefined,
reference_images: referenceImages.length > 0 ? referenceImages : undefined,
});
} catch (error) {
console.error("Together AI API error:", error);
+15 -24
View File
@@ -3,8 +3,7 @@
import { Navbar } from "@/components/landing/navbar"
import { Footer } from "@/components/landing/footer"
import { LandingHero } from "@/components/landing/hero-section"
import { StoryInput } from "@/components/landing/story-input"
import { CreateButton } from "@/components/landing/create-button"
import { ComicCreationForm } from "@/components/landing/comic-creation-form"
import { useState, useEffect } from "react"
export default function Home() {
@@ -43,28 +42,20 @@ export default function Home() {
<div className="max-w-xl mx-auto lg:mx-0 w-full z-10">
<LandingHero />
<div className="space-y-4 sm:space-y-5 mt-4 sm:mt-5">
<div className="opacity-0 animate-fade-in-up animation-delay-100">
<StoryInput
prompt={prompt}
setPrompt={setPrompt}
style={style}
setStyle={setStyle}
characterFiles={characterFiles}
setCharacterFiles={setCharacterFiles}
isLoading={isLoading}
/>
</div>
<div className="opacity-0 animate-fade-in-up animation-delay-200">
<CreateButton
prompt={prompt}
style={style}
characterFiles={characterFiles}
isLoading={isLoading}
setIsLoading={setIsLoading}
/>
</div>
</div>
<div className="space-y-4 sm:space-y-5 mt-4 sm:mt-5">
<div className="opacity-0 animate-fade-in-up animation-delay-100">
<ComicCreationForm
prompt={prompt}
setPrompt={setPrompt}
style={style}
setStyle={setStyle}
characterFiles={characterFiles}
setCharacterFiles={setCharacterFiles}
isLoading={isLoading}
setIsLoading={setIsLoading}
/>
</div>
</div>
</div>
</div>
@@ -1,13 +1,15 @@
"use client";
import { useState } from "react";
import { useRef, useEffect } from "react";
import { Upload, X, Check } from "lucide-react";
import { useState, useRef, useEffect } from "react";
import { useRouter } from "next/navigation";
import { Upload, X, Check, ArrowRight, Loader2 } from "lucide-react";
import { Button } from "@/components/ui/button";
import { useToast } from "@/hooks/use-toast";
import { useS3Upload } from "next-s3-upload";
import { useAuth, SignInButton } from "@clerk/nextjs";
import { COMIC_STYLES } from "@/lib/constants";
interface StoryInputProps {
interface ComicCreationFormProps {
prompt: string;
setPrompt: (prompt: string) => void;
style: string;
@@ -15,9 +17,10 @@ interface StoryInputProps {
characterFiles: File[];
setCharacterFiles: (files: File[]) => void;
isLoading: boolean;
setIsLoading: (loading: boolean) => void;
}
export function StoryInput({
export function ComicCreationForm({
prompt,
setPrompt,
style,
@@ -25,11 +28,32 @@ export function StoryInput({
characterFiles,
setCharacterFiles,
isLoading,
}: StoryInputProps) {
setIsLoading,
}: ComicCreationFormProps) {
const router = useRouter();
const [loadingStep, setLoadingStep] = useState(0);
const { toast } = useToast();
const { uploadToS3 } = useS3Upload();
const { isSignedIn, isLoaded } = useAuth();
const [hasApiKey, setHasApiKey] = useState(false);
const [previews, setPreviews] = useState<string[]>([]);
const [showPreview, setShowPreview] = useState<number | null>(null);
const [showStyleDropdown, setShowStyleDropdown] = useState(false);
const fileInputRef = useRef<HTMLInputElement>(null);
const textareaRef = useRef<HTMLTextAreaElement>(null);
// Check if user has their own API key set
useEffect(() => {
const checkApiKey = () => {
const apiKey = localStorage.getItem("together_api_key");
setHasApiKey(!!apiKey);
};
checkApiKey();
// Listen for storage changes
window.addEventListener("storage", checkApiKey);
return () => window.removeEventListener("storage", checkApiKey);
}, []);
useEffect(() => {
if (isLoading) {
@@ -37,6 +61,35 @@ export function StoryInput({
}
}, [isLoading]);
useEffect(() => {
// Auto-focus the textarea when component mounts
if (textareaRef.current) {
textareaRef.current.focus();
}
}, []);
useEffect(() => {
if (!isLoading) return;
const steps = [
"Enhancing prompt...",
"Generating scenes...",
"Creating your comic...",
];
let currentStep = 0;
const interval = setInterval(() => {
currentStep += 1;
if (currentStep < steps.length) {
setLoadingStep(currentStep);
} else {
clearInterval(interval);
}
}, 2500);
return () => clearInterval(interval);
}, [isLoading]);
const handleFiles = (newFiles: FileList | null) => {
if (!newFiles) return;
@@ -84,6 +137,83 @@ export function StoryInput({
return () => document.removeEventListener("mousedown", handleClickOutside);
}, []);
const handleCreate = async () => {
if (!prompt.trim()) {
toast({
title: "Prompt required",
description: "Please enter a prompt to generate your comic",
variant: "destructive",
duration: 3000,
});
return;
}
setIsLoading(true);
setLoadingStep(0);
try {
const apiKey = localStorage.getItem("together_api_key");
const characterUploads = await Promise.all(
characterFiles.map((file) => uploadToS3(file).then(({ url }) => url))
);
// Use API to create story and generate first page
const response = await fetch("/api/generate-comic", {
method: "POST",
headers: {
"Content-Type": "application/json",
},
body: JSON.stringify({
prompt,
apiKey,
style,
characterImages: characterUploads,
}),
});
if (!response.ok) {
const errorData = await response.json();
if (response.status === 429 && errorData.isRateLimited) {
throw new Error(errorData.error);
}
throw new Error(errorData.error || "Failed to create story");
}
const result = await response.json();
// Redirect to the story editor using slug
router.push(`/editor/${result.storySlug}`);
} catch (error) {
console.error("Error creating comic:", error);
toast({
title: "Creation failed",
description:
error instanceof Error
? error.message
: "Failed to create comic. Please try again.",
variant: "destructive",
duration: 4000,
});
setIsLoading(false);
}
};
const handleKeyDown = (e: React.KeyboardEvent<HTMLTextAreaElement>) => {
const isEnter = e.key === "Enter" || e.key === "\n" || e.keyCode === 13;
const isModifierPressed = e.shiftKey || e.ctrlKey || e.metaKey; // metaKey for Cmd on Mac
if (isEnter && isModifierPressed) {
e.preventDefault();
handleCreate();
}
};
const loadingSteps = [
"Enhancing prompt...",
"Generating scenes...",
"Creating your comic...",
];
return (
<>
<div className="relative glass-panel p-0.5 sm:p-1 rounded-xl group focus-within:border-indigo/30 transition-colors">
@@ -95,8 +225,10 @@ export function StoryInput({
</div>
<textarea
ref={textareaRef}
value={prompt}
onChange={(e) => setPrompt(e.target.value)}
onKeyDown={handleKeyDown}
placeholder="A cyberpunk detective standing in neon rain, holding a glowing datapad, moody lighting, noir style..."
disabled={isLoading}
className="w-full bg-transparent border-none text-sm text-white placeholder-muted-foreground/50 focus:ring-0 focus:outline-none resize-none h-16 leading-relaxed disabled:opacity-50 disabled:cursor-not-allowed"
@@ -242,6 +374,48 @@ export function StoryInput({
</div>
</div>
)}
<div className="pt-2">
{!isLoaded ? (
<div className="h-10" />
) : isSignedIn ? (
<div className="flex items-center justify-between gap-3 w-full">
<Button
onClick={handleCreate}
disabled={isLoading || !prompt.trim()}
className="bg-white hover:bg-neutral-200 text-black px-8 py-2 rounded-md text-sm font-medium transition-colors flex items-center justify-center gap-3 tracking-tight"
>
{isLoading ? (
<>
<Loader2 className="w-4 h-4 animate-spin" />
<span className="text-sm font-medium tracking-tight">
{loadingSteps[loadingStep]}
</span>
</>
) : (
<>
Generate
<ArrowRight className="w-4 h-4" />
</>
)}
</Button>
<div className="text-xs text-muted-foreground whitespace-nowrap">
{hasApiKey ? (
<>Using your API key (~$0.01 per comic)</>
) : (
<>1 credit weekly</>
)}
</div>
</div>
) : (
<SignInButton mode="modal">
<Button className="w-full sm:w-auto sm:min-w-40 bg-white hover:bg-neutral-200 text-black px-8 py-2 rounded-md text-sm font-medium transition-colors flex items-center justify-center gap-3 tracking-tight">
Login to create your comic
<ArrowRight className="w-4 h-4" />
</Button>
</SignInButton>
)}
</div>
</>
);
}
-178
View File
@@ -1,178 +0,0 @@
"use client";
import { useState, useEffect } from "react";
import { useRouter } from "next/navigation";
import { ArrowRight, Loader2 } from "lucide-react";
import { Button } from "@/components/ui/button";
import { useToast } from "@/hooks/use-toast";
import { useS3Upload } from "next-s3-upload";
import { useAuth, SignInButton } from "@clerk/nextjs";
interface CreateButtonProps {
prompt: string;
style: string;
characterFiles: File[];
isLoading: boolean;
setIsLoading: (loading: boolean) => void;
}
export function CreateButton({
prompt,
style,
characterFiles,
isLoading,
setIsLoading,
}: CreateButtonProps) {
const router = useRouter();
const [loadingStep, setLoadingStep] = useState(0);
const { toast } = useToast();
const { uploadToS3 } = useS3Upload();
const { isSignedIn, isLoaded } = useAuth();
const [hasApiKey, setHasApiKey] = useState(false);
// Check if user has their own API key set
useEffect(() => {
const checkApiKey = () => {
const apiKey = localStorage.getItem("together_api_key");
setHasApiKey(!!apiKey);
};
checkApiKey();
// Listen for storage changes
window.addEventListener("storage", checkApiKey);
return () => window.removeEventListener("storage", checkApiKey);
}, []);
useEffect(() => {
if (!isLoading) return;
const steps = [
"Enhancing prompt...",
"Generating scenes...",
"Creating your comic...",
];
let currentStep = 0;
const interval = setInterval(() => {
currentStep += 1;
if (currentStep < steps.length) {
setLoadingStep(currentStep);
} else {
clearInterval(interval);
}
}, 2500);
return () => clearInterval(interval);
}, [isLoading]);
const handleCreate = async () => {
if (!prompt.trim()) {
toast({
title: "Prompt required",
description: "Please enter a prompt to generate your comic",
variant: "destructive",
duration: 3000,
});
return;
}
setIsLoading(true);
setLoadingStep(0);
try {
const apiKey = localStorage.getItem("together_api_key");
const characterUploads = await Promise.all(
characterFiles.map((file) => uploadToS3(file).then(({ url }) => url))
);
// Use API to create story and generate first page
const response = await fetch("/api/generate-comic", {
method: "POST",
headers: {
"Content-Type": "application/json",
},
body: JSON.stringify({
prompt,
apiKey,
style,
characterImages: characterUploads,
}),
});
if (!response.ok) {
const errorData = await response.json();
if (response.status === 429 && errorData.isRateLimited) {
throw new Error(errorData.error);
}
throw new Error(errorData.error || "Failed to create story");
}
const result = await response.json();
// Redirect to the story editor using slug
router.push(`/editor/${result.storySlug}`);
} catch (error) {
console.error("Error creating comic:", error);
toast({
title: "Creation failed",
description:
error instanceof Error
? error.message
: "Failed to create comic. Please try again.",
variant: "destructive",
duration: 4000,
});
setIsLoading(false);
}
};
const loadingSteps = [
"Enhancing prompt...",
"Generating scenes...",
"Creating your comic...",
];
return (
<div className="pt-2">
{!isLoaded ? (
<div className="h-10" />
) : isSignedIn ? (
<div className="flex items-center justify-between gap-3 w-full">
<Button
onClick={handleCreate}
disabled={isLoading || !prompt.trim()}
className="bg-white hover:bg-neutral-200 text-black px-8 py-2 rounded-md text-sm font-medium transition-colors flex items-center justify-center gap-3 tracking-tight"
>
{isLoading ? (
<>
<Loader2 className="w-4 h-4 animate-spin" />
<span className="text-sm font-medium tracking-tight">
{loadingSteps[loadingStep]}
</span>
</>
) : (
<>
Generate
<ArrowRight className="w-4 h-4" />
</>
)}
</Button>
<div className="text-xs text-muted-foreground whitespace-nowrap">
{hasApiKey ? (
<>Using your API key (~$0.01 per comic)</>
) : (
<>1 credit weekly</>
)}
</div>
</div>
) : (
<SignInButton mode="modal">
<Button className="w-full sm:w-auto sm:min-w-40 bg-white hover:bg-neutral-200 text-black px-8 py-2 rounded-md text-sm font-medium transition-colors flex items-center justify-center gap-3 tracking-tight">
Login to create your comic
<ArrowRight className="w-4 h-4" />
</Button>
</SignInButton>
)}
</div>
);
}
+38 -5
View File
@@ -82,13 +82,46 @@ export async function getStoryWithPagesBySlug(slug: string): Promise<{ story: St
}
export async function getStoryCharacterImages(storyId: string): Promise<string[]> {
const storyPages = await db.select({ characterImageUrls: pages.characterImageUrls })
const storyPages = await db.select({
characterImageUrls: pages.characterImageUrls,
pageNumber: pages.pageNumber
})
.from(pages)
.where(eq(pages.storyId, storyId));
.where(eq(pages.storyId, storyId))
.orderBy(pages.pageNumber);
// Flatten all character URLs from all pages and remove duplicates
const allUrls = storyPages.flatMap(page => page.characterImageUrls);
return [...new Set(allUrls)]; // Remove duplicates
// Flatten all character URLs from all pages, keeping order by page number
const allUrls: string[] = [];
const seenUrls = new Set<string>();
for (const page of storyPages) {
for (const url of page.characterImageUrls) {
if (!seenUrls.has(url)) {
seenUrls.add(url);
allUrls.push(url);
}
}
}
return allUrls;
}
export async function getLastPageImage(storyId: string): Promise<string | null> {
const allPages = await db.select({ generatedImageUrl: pages.generatedImageUrl, pageNumber: pages.pageNumber })
.from(pages)
.where(eq(pages.storyId, storyId))
.orderBy(pages.pageNumber);
if (allPages.length === 0) return null;
// Find the last page that has a generated image
for (let i = allPages.length - 1; i >= 0; i--) {
if (allPages[i].generatedImageUrl) {
return allPages[i].generatedImageUrl;
}
}
return null;
}
export async function getNextPageNumber(storyId: string): Promise<number> {