GitHub Repository

AI-powered animated comic generator — transform scripts into fully animated videos with AI-driven character design, storyboarding, and video synthesis.

1,548 starsTypeScript

I built an AI comic generator from scratch using only natural language

Name: I built an AI comic generator from scratch using only natural language
Availability: InStock
Author: twwch

by twwch·Mar 12, 2026·1 point·3 comments

Visit Project View on HN

AI Analysis

●●SolidShip ItRabbit Hole

Four hours of vibe coding with auto-commits as safety nets.

Strengths

•Auto-commits after each feature provide rollback safety when AI breaks things.
•Subagent-driven development with self-review before execution is clever workflow.
•Plain-text requirements to architecture doc to code is streamlined approach.

Weaknesses

•The product itself is AIComicBuilder—this post is just the build story.
•Vibe coding workflows are becoming common; not a unique differentiator.

Post Description

I spent ~4 hours building AIComicBuilder [1], a full-stack AI comic/drama generation platform, without writing a single line of code manually. This is a writeup of my "vibe coding" workflow using Claude Code.

The app lets you: input a script → AI generates screenplay → character analysis → storyboard generation (with first/last frame images) → video generation. Supports OpenAI-compatible, Gemini, and Bytedance Seedance APIs for text/image/video models, configurable per-user via a settings UI.

Tech: Next.js full-stack, SQLite, queue-based async processing.

**The workflow:** 1. Describe tech requirements in plain text → AI generates architecture doc → review/tweak in the terminal 2. Reference the doc, load `superpowers:subagent-driven-development` → AI writes the plan, self-reviews it, then executes autonomously (~45 min) 3. Auto-commits after each feature (critical safety net if AI breaks something) 4. Bug? Screenshot + one sentence → AI fixes it 5. Ugly UI? "redesign using ui-ux-pro-max" → 3 iterations → done

The key insight from Karpathy's "vibe coding" concept: in this paradigm, *documentation quality matters more than coding ability*. The clearer your requirements, the better the output. Your role shifts from programmer to product manager.

I'm documenting the full workflow and useful Claude Code skills here: [2]

Curious if others have found good skill combinations for Claude Code, or are using similar approaches with other tools (Cursor, Aider, etc.).

[1] https://github.com/twwch/AIComicBuilder [2] https://github.com/twwch/vibe-coding