Open Genmoji aims to replicate Apple's Genmoji feature using open-source technologies, enabling users to generate custom emojis from text prompts. It's designed for users familiar with image generation models and those who want to create and use personalized emojis across various platforms, including iOS.

How It Works

The project leverages LoRA (Low-Rank Adaptation) files, finetuned on thousands of Apple emojis, to adapt existing image generation models to produce emoji outputs. It also provides "Prompt Assist" metaprompts to help users craft effective prompts for the models. This approach allows for flexibility, as the LoRAs can be applied to any compatible image generation model.

Quick Start & Requirements

• Installation: Clone the repository, create a Python 3.11 virtual environment, install huggingface_hub and mflux (pip install -U huggingface_hub mflux), and run python3 download.py to select and download LoRA weights.
• Prerequisites: Python 3.11 is strongly recommended. Hugging Face CLI login and agreement to gated model terms (for Flux.1 Dev) are required. The mflux tool is specified for macOS.
• Resources: LoRA files are approximately 209MB.
• Links: HuggingFace, Quickstart, Tutorial

Highlighted Details

• Recreates Apple's Genmoji functionality with open technology.
• LoRAs are available on HuggingFace for various emoji types and models.
• Includes "Prompt Assist" metaprompts to improve prompt quality for emoji generation.
• Generated emojis can be used as stickers on iOS 17+ and as inline emojis on iOS 18+.
• A Python workflow is provided to chain LM Studio (for LLM prompts) and mflux for automated generation.

Maintenance & Community

• The project is maintained by EvanZhouDev.
• Information on contributing to finetuning and adding new LoRAs is available.

Licensing & Compatibility

• The README does not explicitly state a license. Compatibility for commercial use or closed-source linking is not specified.

Limitations & Caveats

• The provided tutorial is macOS-specific, though the project is adaptable to other operating systems with alternative image generation tools.
• The project relies on specific versions of Python and potentially gated models, which may require careful setup.