MimicThem addresses the desire to replicate popular Xiaohongshu (Little Red Book) image-text post aesthetics. It provides a one-click solution for users to input a Xiaohongshu URL and generate visually similar content using AI. The tool automates prompt inference, image generation with variations, and copywriting, offering a streamlined workflow for content creators seeking trending styles.

How It Works

The project leverages a Python backend (FastAPI) and a Next.js frontend. Upon receiving a Xiaohongshu URL, it downloads original images and employs the ByteDance Seed 2.0 Pro model to reverse-engineer descriptive prompts. These prompts then feed into the Seedream 4.5 model for initial image generation. Subsequently, multiple prompt variations are created, and a text-to-image model generates diverse image variants, maintaining character consistency. The system also generates Xiaohongshu-style captions and tags, with results pushed via Server-Sent Events (SSE) for real-time user feedback. Parallel processing optimizes download and variant generation efficiency.

Quick Start & Requirements

• Installation: Requires setting up both a Python backend (pip install -r requirements.txt within backend/) and a Node.js frontend (npm install within frontend/). Backend is launched with python main.py, frontend with npm run dev.
• Prerequisites:
  • Volcano Engine ARK API Key.
  • Access to doubao-seed-2-0-pro-260215 and doubao-seedream-4-5-251128 models.
  • Python 3.12+.
  • Node.js environment.
• Endpoints: Backend typically runs on http://localhost:8000, frontend on http://localhost:3000.
• Resources: Configuration involves editing backend/config.ini with the ARK API Key.
• Docs: Bilibili introduction video available at https://b23.tv/ag2eCca.

Highlighted Details

• One-Click Replication: Simplifies the process by accepting a single Xiaohongshu URL.
• Intelligent Prompt Inference: Automatically derives prompts from input images.
• AI-Powered Image Generation: Utilizes Seedream and text-to-image models to create base images and multiple stylistic variants.
• Automated Copywriting: Generates relevant, Xiaohongshu-style captions and hashtags.
• Real-time Processing Feedback: Employs SSE for live updates on generation progress and immediate display of results.
• Optimized Workflow: Parallel execution enhances the speed of image downloading and variant generation.

Maintenance & Community

No specific details regarding maintainers, community channels (e.g., Discord, Slack), or project roadmap are provided in the README.

Licensing & Compatibility

• License: MIT License.
• Compatibility: The project explicitly states it is for "learning and research use only." Generated content is restricted from commercial use or copyright infringement, with users assuming all responsibility. This effectively limits commercial application.

Limitations & Caveats

• Content Type: Strictly supports only image-text Xiaohongshu posts; video content is not processed.
• Output Nature: Generated images are AI-created interpretations, not exact duplicates of original assets.
• Dependency: Requires configuration of specific third-party API keys (Volcano Engine ARK) and model access, posing an adoption barrier.
• Usage Restriction: The MIT license is accompanied by a disclaimer prohibiting commercial use of generated outputs.