Hchen1218/heytea-style

Make your photo look like a Heytea style poster

What it solves

This project provides a creative toolkit for transforming real-life photos into stylized visual content. It solves the problem of creating consistent, character-based marketing-style posters and interactive desktop pets from a single source image, automating the process of extracting visual elements (like colors and textures) to generate new artistic interpretations.

How it works

The system operates as a "Skill" within the Codex environment, using a multi-step confirmation workflow to ensure visual consistency:

  1. Visual Extraction: It analyzes a photo to identify the main subject and its characteristics (colors, materials, and components).
  2. cantidad Generation: Depending on the user's choice, it generates either a typography-focused poster, a narrative doodle poster, or a "Flavor Monster" character based on the extracted visual data.
  3. Character Locking: Once a "Flavor Monster" is confirmed, the system locks its identity (body, face, limbs, colors) to maintain consistency across different outputs.
  4. Deployment: For interactive elements, it uses an Electron-based runtime to deploy the character as a transparent, overlaying desktop pet for macOS or Windows, supporting various animations and physics-based behaviors.

Who it’s for

  • Content Creators: Those looking to create unique, hand-drawn style posters from product photos.
  • Visual Designers: Users wanting to experiment with character design based on product aesthetics.
  • Hobbyists: People who want personalized, interactive AI-generated characters on their desktops.

Highlights

  • Consistent Character Identity: Locks character traits to ensure the same monster appears across posters and desktop pets.
  • Multi-format Output: Generates typography posters, narrative doodles, and interactive software.
  • Crossover Interaction: Can place the generated character back into the original photo with specific actions (e.g., pushing or tasting) to create a story.
  • Cross-platform Desktop Pets: Includes a dedicated Electron runtime for running characters on macOS and Windows with support for physics and complex animations.

Related

  • Project
    AMAP-ML/DreamX-Creator

    A research framework for native joint audio-video generation that produces synchronized, high-quality 2K resolution content from a first frame and a text prompt.

  • Project
    openaiotlab/CUHK-X

    A large-scale multimodal dataset and benchmark for human activity recognition and reasoning, integrating seven synchronized sensor modalities to enable complex action understanding.

  • Project
    TetreesEX/TetreesAgent_EX

    Tetrees Agent EX is a Node.js SDK/CLI that lets developers search, quote, run, extend, and publish AI Packs on the Tetrees EX marketplace via the public MCP contract. It provides example scripts for buyer, builder, and seller flows, and keeps all credentials local.

  • Project
    huggingface/huggingface-gemma-recipes

    A collection of minimal recipes and notebooks for implementing multimodal inference, fine-tuning, and RAG with the Gemma family of models.