DEV Community: Junho

We Built ComfyUI Workflow Visualization Into Our AI Art Portfolio Platform

Junho — Thu, 19 Mar 2026 15:35:34 +0000

Hey dev community! 👋

I'm building PixelAI — a portfolio platform for AI artists. Today I shipped a feature I'm really excited about: ComfyUI workflow support with interactive graph visualization.

The Problem

ComfyUI is arguably the most powerful tool for AI image/video generation. But sharing workflows is painful:

Screenshot your node graph
Paste prompts in Discord
Upload .json files to random file hosts
Explain your setup in comment threads

The Solution

Now on PixelAI, you can:

Upload your .json workflow alongside your artwork
Auto-extraction — checkpoint, sampler, seed, steps, CFG are pulled from the workflow automatically
Interactive node graph — built with React Flow, viewers can zoom, pan, and explore your workflow
One-click download — viewers download your exact .json and load it in ComfyUI

Technical Implementation

Parser: Supports both Workflow format (File → Save) and API format (File → Export API)
Visualization: React Flow with lazy loading (dynamic import, zero bundle impact on other pages)
Node categories: Color-coded — Loader (purple), Sampler (blue), Conditioning (yellow), Output (red)
Custom nodes: Detected and marked with ⚡ badge
Edge labels: Show data types (IMAGE, VIDEO, MASK, etc.)
Fullscreen mode: For complex workflows with 50+ nodes
Validation: Non-ComfyUI JSON files are rejected with helpful error message

Stack

Next.js 14 (App Router)
React Flow (@xyflow/react)
Supabase (jsonb column for workflow storage)
Tailwind CSS

What's Next

A/B comparison of different workflows for the same image
Workflow versioning (iterate and track changes)
"Fork" button — remix someone's workflow with your modifications

Check it out: pixelai.world

How We're Building the Portfolio Platform AI Artists Deserve

Junho — Tue, 17 Mar 2026 15:02:51 +0000

We Built GitHub for AI Art — Here's What We Shipped This Week
AI art is booming. Millions create with Midjourney, Stable Diffusion, DALL·E, and Flux every day. But where does that art live?

Instagram wasn't built for AI art. Twitter compresses your images. DeviantArt feels stuck in 2010.

We're building PixelAI — a portfolio-first platform for AI artists. Think GitHub, but for generative art.

This week we shipped a massive update. Here's everything that's new.

📂 Collections
Every great portfolio tells a story. Collections let you group works into themed projects — "Neon Dreams", "Character Studies", "Client Work" — whatever makes sense for your creative process.

Public or private
Custom descriptions
Grid view with thumbnails
Add works from any post detail page
🏷️ Tags & Full-Text Search
10 genre categories: Character, Landscape, Abstract, Portrait, Concept Art, Photography, Illustration, 3D Render, Logo & Icon, and Anime & Manga.

Every post is searchable by title, description, and prompt text. Combined with genre chips and model filters, discovery just works.

⚙️ Generation Parameters
The AI art community runs on reproducibility. Now you can share:

Negative Prompt
Seed, Steps, CFG Scale, Sampler
Up to 5 custom key-value parameters
Progressive disclosure — the upload form stays clean for beginners, power users expand "Advanced Parameters" for full control.

🖼️ Multi-Image Carousel
Upload up to 10 images per post. Swipe on mobile, arrows on desktop. Perfect for process shots, model comparisons, prompt variations.

⚡ 2x Faster Page Loads
We rewrote every server-side page to use parallel data fetching. 7-8 sequential DB queries → 2-3 parallel batches. Portfolio page: ~1.5s → under 600ms.

🔜 What's Next
Bulk Upload — Upload 20+ works at once
Dynamic OG Images — Beautiful social cards
Product Hunt Launch — Coming soon 👀
Try it free: pixelai.world

Your work deserves a proper home. Not a social media feed. A portfolio.

I Built an AI Art Community as a Solo Dev — Here's How

Junho — Thu, 12 Mar 2026 15:58:49 +0000

AI image generation has exploded. Midjourney, Stable Diffusion, DALL-E, Flux — everyone's creating incredible images. But where do you share them?

Instagram isn't built for AI art. Twitter buries your work in the feed. Discord servers are chaotic. I wanted a dedicated space where AI artists could share work, discover prompts, and actually connect with each other.

So I built PixelAI.

The Problem

I noticed a gap: AI image creators had amazing tools but no real community platform. They were scattered across Discord servers, Reddit threads, and Twitter hashtags. There was no central place to:

Share AI-generated artwork with proper attribution to the model used
Discover how people made their images (prompts are gold)
Compete in creative challenges
Build a portfolio of AI art

Tech Stack

I went with what I know and what ships fast:

Next.js 15 (App Router) — React Server Components for fast initial loads
Supabase — Auth, Postgres database, file storage. The whole backend in one service
shadcn/ui — Clean, accessible components without fighting a design system
Vercel — Deploy on push, zero config
TypeScript — Because debugging runtime type errors at 2 AM isn't fun

Why This Stack?

Speed to ship. As a solo dev, I needed to go from idea to production fast. Supabase eliminated the need to build auth, set up a database, or configure file storage separately. Next.js App Router with Server Components meant I could fetch data on the server without building a separate API layer.

Total time from npx create-next-app to production: about 2 weeks of evening/weekend work.

Key Features

AI Model Tagging
Every post is tagged with the AI model used — Midjourney, Stable Diffusion, DALL-E, Flux, ComfyUI, and more. You can filter and discover art by model.
Prompt Sharing
This is the killer feature. Creators can optionally share their prompts. Good prompts are incredibly valuable, and having them attached to the actual output makes learning so much easier.
Weekly Challenges
Every week, there's a themed challenge. Users submit entries, the community votes, and winners get featured. It drives engagement and gives people a reason to come back.
Social Features
Follow creators, like posts, comment, share. The basics — but done right for AI art specifically.

Architecture Decisions

Server Components First
Most pages are Server Components that fetch data directly from Supabase. Client Components only where needed (interactions, real-time updates). This keeps the bundle small and initial loads fast.

// Server Component — no client JS shipped
export default async function PostPage({ params }) {
const supabase = await createClient();
const { data: post } = await supabase
.from("posts")
.select(", user:users()")
.eq("id", params.id)
.single();

return ;
}

Row Level Security
Supabase RLS means I don't need a middleware layer for authorization. The database enforces who can read/write what:

-- Users can only delete their own posts
CREATE POLICY "delete_own_posts" ON posts
FOR DELETE USING (auth.uid() = user_id);

Image Handling
All images go through Supabase Storage with public buckets. Next.js `` handles optimization and responsive sizes automatically. No need for a separate CDN setup.

Challenges I Faced

Feed Algorithm
Simple reverse-chronological wasn't enough. If you follow people, you want their posts first, then discover new content. I built a two-phase feed: following posts first, then seamlessly transitioning to latest posts — with deduplication and scroll position restoration.
Prompt Privacy During Challenges
During active challenges, prompts need to be hidden (otherwise people just copy the winning formula). But after the challenge ends, winner prompts should be revealed. This required careful RLS policies and state management.
Content Moderation
AI can generate... anything. For now, I'm relying on community reporting and manual review. It's not scalable, but it works at the current size.

What I Learned

Ship early, iterate fast. The first version was embarrassingly simple. But real users give you better feedback than hypothetical planning.
Supabase is incredible for solo devs. Auth, database, storage, real-time — all in one. The time saved is massive.
Seed your own content. An empty community is a dead community. I uploaded 50+ posts before telling anyone about it.
Prompts are the real value. People don't just want to see AI art — they want to learn how to make it. Prompt sharing drives the most engagement.

What's Next

Better discovery — recommendation engine based on likes and follows
Collections — let users curate themed galleries
Mobile app — the web app is responsive, but native would be better

Try It Out

If you're into AI art (creating or just appreciating), check out pixelai.world. It's free, and I'd love your feedback.

The community is still early, which means your voice actually matters in shaping what it becomes. Drop a comment here or reach out on the site — I read everything.

Built by a solo dev who believes AI art deserves its own home. 🎨