1. Introduction
The facechain character photo application has open sourced the first version of ID photo generation since August 11th. Currently, there are 6K+ stars on github (GitHub - modelscope/facechain: FaceChain is a deep-learning toolchain for generating your Digital-Twin.). It aims to create a character-centered AIGC toolbox. Currently, its character photo function has rich online experience forms and entrances. The main ones are as follows:
a.) Wanxiang Photo Gallery online experience:Tongyi Wanxiang
b.) FaceChain Lingji commercial API:facechain, portrait photo_Lingji model service-Alibaba Cloud Help Center
c.) Magic character photo online experience: FaceChain character photo generation
In addition to the above experience entrance, this project can also be integrated as a plug-in in sd webui, and is also being expanded simultaneously in the plug-in of comfyui. At the functional level, it currently has functions such as character portraits, virtual fittings, sad talker, etc., and more applications are being developed.
论文链接：FaceChain: A Playground for Identity-Preserving Portrait Generation：https://arxiv.org/abs/2308.14256。
The main functions of facechain character photos are shown in the figure below:
Recent active updates support many new features. Below we have excerpted some of the core functions for display.
2. Core functions
1. Unlimited photo styles
The most surprising thing is that it currently supports hundreds of photo styles. And the styles are still being continuously updated, some styles are as follows:
Not only can you use pre-made photo styles, you can also upload your favorite photo style LoRA locally (github submissions for PR contributions are also welcome),
After the upload is successful (as shown in the picture above fire_dress), enter the prompt word and click Generate to display the customized style picture:
2. Group photo with multiple people
In addition to single-person portraits, facechain currently also supports the function of multi-person portraits. The renderings are as follows:
3. The posture can be specified
The designated posture effect of outdoor wedding dress is as follows:
4. Templates can be specified
When you encounter a background you like, you can replace only the head part and perfectly retain the background. The interactive interface is as follows:
The renderings are as follows:
3. Open source platform
1、GitHub：
2. Modelscope project base camp:
FaceChain portrait photography
3、huggingface space
demo体验：https://huggingface.co/spaces/modelscope/FaceChain
4. Community Ecology
1. Build projects together with developers on DingTalk
2. Contributor incentives
a.) GPU duration: Contributors to the project can also independently declare their contributions to redeem the GPU rewards on the Magic Platform: Developer Contribution Record
Submissions for tutorials and other articles can be sent to: [email protected].
b.) Certificate of Honor: For developers who have made significant contributions, we will also award certificates on the Honor Wall. The rendering is as follows:
3. Global developer invitation
If you agree with our project convention: interesting, useful, good, and beautiful, please sign up immediately to join! We look forward to the continued development of the FaceChain open source project and leading the new trend of artificial intelligence technology.
4. Future planning
At present, the overall planning of facechain mainly has four dimensions: a.) real-life photo style, b.) virtual photo style, c.) expanded applications such as virtual fitting, d.) ecological plug-ins such as sd webui. Details are as follows:
