[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-72270":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":17,"stars30d":18,"stars90d":16,"forks30d":16,"starsTrendScore":18,"compositeScore":19,"rankGlobal":10,"rankLanguage":10,"license":20,"archived":21,"fork":21,"defaultBranch":22,"hasWiki":21,"hasPages":21,"topics":23,"createdAt":10,"pushedAt":10,"updatedAt":24,"readmeContent":25,"aiSummary":26,"trendingCount":16,"starSnapshotCount":16,"syncStatus":27,"lastSyncTime":28,"discoverSource":29},72270,"PuLID","ToTheBeginning\u002FPuLID","ToTheBeginning","[NeurIPS 2024] Official code for PuLID: Pure and Lightning ID Customization via Contrastive Alignment","",null,"Python",3544,267,56,157,0,3,9,67.68,"Apache License 2.0",false,"main",[],"2026-06-12 04:01:04","# PuLID (NeurIPS 2024)\n\n### :open_book: PuLID: Pure and Lightning ID Customization via Contrastive Alignment\n> [![arXiv](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FarXiv-Paper-\u003CCOLOR>.svg)](https:\u002F\u002Farxiv.org\u002Fabs\u002F2404.16022) [![xl](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F🤗-HuggingFaceDemo-orange)](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002Fyanze\u002FPuLID) [![flux](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F🤗-PuLID_FLUX_demo-orange)](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002Fyanze\u002FPuLID-FLUX) [![Replicate](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FReplicate-Demo_for_PuLID-blue)](https:\u002F\u002Freplicate.com\u002Fzsxkib\u002Fpulid) [![Replicate](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FReplicate-PuLID_FLUX-blue)](https:\u002F\u002Freplicate.com\u002Fzsxkib\u002Fflux-pulid)\u003Cbr>\n> Zinan Guo*, Yanze Wu*✝, Zhuowei Chen, Lang Chen, Peng Zhang, Qian He \u003Cbr>\n> (*Equal Contribution, ✝Corresponding Author) \u003Cbr>\n> ByteDance Inc \u003Cbr>\n\n🔥🔥 We are excited to share [DreamO](https:\u002F\u002Fgithub.com\u002Fbytedance\u002FDreamO), a unified image customization framework that supports ID, IP, Try-On, and style tasks. Check it out and see what it can do! 🔥🔥\n\n### :triangular_flag_on_post: Updates\n* **2024.10.31**: 🔥 We are happy to release our latest [models](https:\u002F\u002Fhuggingface.co\u002Fguozinan\u002FPuLID), **PuLID-v1.1** and **PuLID-FLUX-v0.9.1**. See more in [Model Zoo](#european_castle-model-zoo) and [pulid v1.1 model](docs\u002Fpulid_v1.1.md). We also update a new revision for the [arXiv paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2404.16022), which includes more results, details, and analysis, please check it out.\n* **2024.09.26**: 🎉 PuLID accepted by NeurIPS 2024\n* **2024.09.12**: We're thrilled to announce the release of the **PuLID-FLUX-v0.9.0 model**. Enjoy exploring its capabilities! 😊 [Learn more about this model](docs\u002Fpulid_for_flux.md)\n* **2024.05.23**: share the [preview of our upcoming v1.1 model](docs\u002Fv1.1_preview.md), please stay tuned\n* **2024.05.01**: release v1 codes&models, also the [🤗HuggingFace Demo](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002Fyanze\u002FPuLID)\n* **2024.04.25**: release arXiv paper.\n\n\n## PuLID for FLUX\nPlease check the doc and demo of PuLID-FLUX [here](docs\u002Fpulid_for_flux.md).\n\n### updates\n- [x] Local gradio demo is ready now\n- [x] Online HuggingFace demo is ready now [![flux](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F🤗-PuLID_FLUX_demo-orange)](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002Fyanze\u002FPuLID-FLUX)\n- [x] We have optimized the codes to support consumer-grade GPUS, and now **PuLID-FLUX can run on a 16GB graphic card**. Check the details [here](https:\u002F\u002Fgithub.com\u002FToTheBeginning\u002FPuLID\u002Fblob\u002Fmain\u002Fdocs\u002Fpulid_for_flux.md#local-gradio-demo)\n- [x] (Community Implementation) Online Replicate demo is ready now [![Replicate](https:\u002F\u002Freplicate.com\u002Fzsxkib\u002Fflux-pulid\u002Fbadge)](https:\u002F\u002Freplicate.com\u002Fzsxkib\u002Fflux-pulid)\n- [x] Local gradio demo supports 12GB graphic card now\n- [x] v0.9.1 is ready now\n\n\nBelow results are generated with PuLID-FLUX.\n![pulid_flux_results](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F7eafb90a-fdd1-4ae7-bc41-8c428d568848)\n\n\n## Examples\nImages generated with our PuLID\n![examples](https:\u002F\u002Fgithub.com\u002FToTheBeginning\u002FPuLID\u002Fassets\u002F11482921\u002F65610b0d-ba4f-4dc3-a74d-bd60f8f5ce37)\nApplications\n\nhttps:\u002F\u002Fgithub.com\u002FToTheBeginning\u002FPuLID\u002Fassets\u002F11482921\u002F9bdd0c8a-99e8-4eab-ab9e-39bf796cc6b8\n\n## :european_castle: Model Zoo\n\n|                                              Version                                               | Base Model |                                                                                              Description                                                                                              |\n|:--------------------------------------------------------------------------------------------------:|:----------:|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------:|\n|              [PuLID-v1](https:\u002F\u002Fhuggingface.co\u002Fguozinan\u002FPuLID\u002Fblob\u002Fmain\u002Fpulid_v1.bin)              |    SDXL    |                                                                                             Paper model.                                                                                              |\n|        [PuLID-v1.1](https:\u002F\u002Fhuggingface.co\u002Fguozinan\u002FPuLID\u002Fblob\u002Fmain\u002Fpulid_v1.1.safetensors)        |    SDXL    |                                                     Compared to PuLID-v1, better compatibility, editability, facial naturalness, and similarity.                                                      |\n| [PuLID-FLUX-v0.9.0](https:\u002F\u002Fhuggingface.co\u002Fguozinan\u002FPuLID\u002Fblob\u002Fmain\u002Fpulid_flux_v0.9.0.safetensors) |    FLUX    |             Our first version for PuLID-FLUX, better prompt-following and image quality (since FLUX is more powerful than SDXL). But ID fidelity is not high enough for some male inputs              |\n| [PuLID-FLUX-v0.9.1](https:\u002F\u002Fhuggingface.co\u002Fguozinan\u002FPuLID\u002Fblob\u002Fmain\u002Fpulid_flux_v0.9.1.safetensors) |    FLUX    | Compared to PuLID-FLUX-v0.9.0, better ID fidelity. From the quantitative metric of ID similarity, the improvement is about 5 percentage points. Meanwhile, the editability remains similar as before. |\n\n\n## :wrench: Dependencies and Installation\n- Python >= 3.9 (Recommend to use [Anaconda](https:\u002F\u002Fwww.anaconda.com\u002Fdownload\u002F#linux) or [Miniconda](https:\u002F\u002Fdocs.conda.io\u002Fen\u002Flatest\u002Fminiconda.html))\n- [PyTorch >= 2.0](https:\u002F\u002Fpytorch.org\u002F) if you don't need flux-dev-fp8, otherwise [PyTorch >= 2.4.1](https:\u002F\u002Fpytorch.org\u002F)\n```bash\n# clone PuLID repo\ngit clone https:\u002F\u002Fgithub.com\u002FToTheBeginning\u002FPuLID.git\ncd PuLID\n# create conda env\nconda create --name pulid python=3.10\n# activate env\nconda activate pulid\n# Install dependent packages\n# 1. if you don't need flux-fp8, e.g., you are using xl or flux-bf16, install the following requirements.txt\npip install -r requirements.txt\n# 2. if you need flux-fp8 (to put flux on consumer-grade gpu), install the following requirements_fp8.txt\npip install -r requirements_fp8.txt\n```\n\n## :zap: Quick Inference\n### Local Gradio Demo\n```bash\n# for v1 version\npython app.py\n\n# for v1.1 version\npython app_v1.1.py --base BASE_MODEL\nUsage:\n  -base: can be RunDiffusion\u002FJuggernaut-XL-v9 or Lykon\u002Fdreamshaper-xl-lightning\n```\n\n### Online HuggingFace Demo\nThanks for the GPU grant from HuggingFace team, you can try PuLID HF demo in \n- [https:\u002F\u002Fhuggingface.co\u002Fspaces\u002Fyanze\u002FPuLID](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002Fyanze\u002FPuLID) for SDXL\n- [https:\u002F\u002Fhuggingface.co\u002Fspaces\u002Fyanze\u002FPuLID-FLUX](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002Fyanze\u002FPuLID-FLUX) for FLUX\n\n## :paperclip: Related Resources\nFollowing are some third-party implementations of PuLID we have found in the Internet. \nWe appreciate the efforts of the respective developers for making PuLID accessible to a wider audience.\nIf there are any PuLID based resources and applications that we have not mentioned here, please let us know, \nand we will include them in this list.\n\n#### Online Demo\n- **Colab**: https:\u002F\u002Fgithub.com\u002Fcamenduru\u002FPuLID-jupyter provided by [camenduru](https:\u002F\u002Fgithub.com\u002Fcamenduru)\n- **Replicate (PuLID)**: https:\u002F\u002Freplicate.com\u002Fzsxkib\u002Fpulid provided by [zsxkib](https:\u002F\u002Fgithub.com\u002Fzsxkib)\n- **Replicate (PuLID-FLUX)**: https:\u002F\u002Freplicate.com\u002Fzsxkib\u002Fflux-pulid provided by [zsxkib](https:\u002F\u002Fgithub.com\u002Fzsxkib)\n\n#### ComfyUI\n- https:\u002F\u002Fgithub.com\u002Fcubiq\u002FPuLID_ComfyUI provided by [cubiq](https:\u002F\u002Fgithub.com\u002Fcubiq), native ComfyUI implementation\n- https:\u002F\u002Fgithub.com\u002FZHO-ZHO-ZHO\u002FComfyUI-PuLID-ZHO provided by [ZHO](https:\u002F\u002Fgithub.com\u002FZHO-ZHO-ZHO), diffusers-based implementation\n\n#### WebUI\n- [SD.Next](https:\u002F\u002Fgithub.com\u002Fvladmandic\u002Fautomatic\u002Fblob\u002Fmaster\u002FCHANGELOG.md#update-for-2024-11-21) Implementation provided by [vladmandic](https:\u002F\u002Fgithub.com\u002Fvladmandic)\n- https:\u002F\u002Fgithub.com\u002FMikubill\u002Fsd-webui-controlnet\u002Fpull\u002F2838 provided by [huchenlei](https:\u002F\u002Fgithub.com\u002Fhuchenlei)\n\n#### Other Applications\n- PuLID-FLUX multi-person generation with [Regional-Prompting-FLUX](https:\u002F\u002Fgithub.com\u002FinstantX-research\u002FRegional-Prompting-FLUX), provided by [Anthony](https:\u002F\u002Fgithub.com\u002Fantonioo-c)\n\n## Disclaimer\nThis project strives to impact the domain of AI-driven image generation positively. Users are granted the freedom to \ncreate images using this tool, but they are expected to comply with local laws and utilize it responsibly. \nThe developers do not assume any responsibility for potential misuse by users.\n\n\n##  Citation\nIf PuLID is helpful, please help to ⭐ the repo.\n\nIf you find this project useful for your research, please consider citing our paper:\n```bibtex\n@InProceedings{guo2024pulid,\n  title={PuLID: Pure and Lightning ID Customization via Contrastive Alignment},\n  author={Guo, Zinan and Wu, Yanze and Chen, Zhuowei and Chen, Lang and Zhang, Peng and He, Qian},\n  booktitle={Advances in Neural Information Processing Systems},\n  year={2024}\n}\n```\n\n## Acknowledgement\nWe would like to acknowledge our core team members, Pengze Zhang and Songtao Zhao, for their significant contributions \nto the continued development and maintenance of the PuLID model. Although they were not listed as authors on the first \nversion of this paper, their efforts played a crucial role in enhancing the model.\n\n## :e-mail: Contact\nIf you have any comments or questions, please [open a new issue](https:\u002F\u002Fgithub.com\u002FToTheBeginning\u002FPuLID\u002Fissues\u002Fnew\u002Fchoose) or feel free to contact [Yanze Wu](https:\u002F\u002Ftothebeginning.github.io\u002F) and [Zinan Guo](mailto:guozinan.1@bytedance.com).","PuLID 是一个通过对比对齐实现纯净且快速的身份定制的项目。该项目利用对比学习技术，能够高效地生成个性化图像，支持用户自定义身份特征。其核心功能包括基于对比学习的身份特征提取与合成，以及针对不同硬件环境（如消费级GPU）的优化。特别适合需要进行图像个性化处理的应用场景，例如虚拟角色创建、个性化头像生成等。此外，项目提供了详细的文档和多个在线演示平台，方便开发者和研究人员快速上手使用。",2,"2026-06-11 03:41:07","high_star"]