zsxkib / flash-face

FlashFace: Human Image Personalization with High-fidelity Identity Preservation

  • Public
  • 1.1K runs
  • GitHub
  • Paper
  • License
Run with an API
Playground API Examples README Versions

Input

Output

Examples

View more examples

Run time and cost

This model runs on Nvidia A40 (Large) GPU hardware. Predictions typically complete within 147 seconds. The predict time for this model varies significantly based on the inputs.

Readme

FlashFace: Human Image Personalization with High-fidelity Identity Preservation

Model Description 🎭

FlashFace is a powerful model for human image personalization that allows users to create custom images within seconds using one or several reference faces. The model excels at preserving the identity of the individuals, even for non-celebrities, while closely following language prompts. FlashFace offers flexibility in adjusting the strength of identity image control and language prompt control, enabling users to fine-tune the generated images to their preferences.

License 📜

FlashFace is released under the MIT License, which means it is open-source and free to use for both personal and commercial purposes. However, please note that the developers of FlashFace or Replicate are not liable for any misuse or damages caused by the use of this model.

It’s important to understand that FlashFace is a distinct model and does not use the InsightFace Swapper models (inswapper128, inswapper256, or inswapper512), which have non-commercial restrictions.

If you intend to use FlashFace for commercial projects, you are free to do so under the terms of the MIT License. However, ensure that your use case complies with all relevant laws, regulations, and ethical guidelines. 🚦 If you have any doubts about the legality or appropriateness of your intended use, please consult with legal professionals. 🧑‍⚖️

Remember, as an open-source project, FlashFace relies on the community to use it responsibly and contribute to its development. If you encounter any issues, have suggestions for improvements, or want to contribute to the project, please visit the FlashFace GitHub repository. 👨‍💻

By using FlashFace, you agree to the terms of the MIT License and acknowledge that the model is provided “as is,” without any warranty or liability from the developers. 🛡️

Usage 🛠️

To use FlashFace on the Replicate WebUI, follow these simple steps:

  1. Upload one or several reference faces of the individual you wish to personalize. 📤

  2. Provide a language prompt describing the desired modifications or context for the generated image. 💬

  3. Adjust the identity image control and language prompt control sliders to achieve the desired balance. 🎚️

  4. Click the “Generate” button and wait for FlashFace to create your personalized image. 🖼️

Caveats and Limitations 🚫

While FlashFace demonstrates remarkable capabilities in human image personalization, it is essential for users to keep the following in mind:

  • The model’s performance may vary depending on the quality and clarity of the reference faces provided. 🖼️
  • Extreme modifications or highly abstract language prompts may result in less accurate or coherent generated images. 🎨
  • FlashFace is a powerful tool that can generate highly realistic images resembling target individuals. However, these images are entirely AI-generated and should not be considered genuine photographs or used to misrepresent or deceive others. 🤖
  • Users have the sole responsibility to use FlashFace ethically and in compliance with all applicable laws and regulations. The generated images should not be used for any harmful, malicious, or illegal purposes, such as harassment, defamation, or creating explicit content without consent. ⚖️
  • The developers of FlashFace or Replicate do not condone or support any misuse of the model and are not liable for any consequences resulting from the actions of individual users. By using FlashFace, users agree to take full responsibility for their use of the model and the generated images. 👥
  • If you are unsure about the legality or ethical implications of your intended use case, please consult with legal professionals and carefully consider the potential impact on others before using FlashFace. 🤔

Remember, while FlashFace opens up exciting possibilities for creative expression and personalization, it is crucial to use the model responsibly and with respect for others. 🌟

Citation 📚

If you find FlashFace useful for your research or applications, please cite the project using the following BibTeX: 📄

@misc{zhang2024flashface,
      title={FlashFace: Human Image Personalization with High-fidelity Identity Preservation},
      author={Shilong Zhang and Lianghua Huang and Xi Chen and Yifei Zhang and Zhi-Fan Wu and Yutong Feng and Wei Wang and Yujun Shen and Yu Liu and Ping Luo},
      year={2024},
      eprint={2403.17008},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

FlashFace was inspired by various related demos and repositories, including IP-Adapter, multimodalart/Ip-Adapter-FaceID, FastComposer, T2I-Adapter, PhotoMaker, and InstantID. We thank them for their great work! 👏