| --- |
| license: apache-2.0 |
| pipeline_tag: text-to-image |
| tags: |
| - diffusion |
| - sdxl |
| - pulid |
| - kinship-image-generation |
| --- |
| |
| # KinDiffusion |
|
|
| Official inference checkpoints for [KinDiffusion](https://github.com/Unilinear/KinDiffusion), a diffusion model for generating child images from two parent photos. |
|
|
| ## Checkpoints |
|
|
| | File | Purpose | Size | |
| | --- | --- | ---: | |
| | `pytorch_model_3.bin` | Identity encoder | 221 MB | |
| | `pytorch_model_4.bin` | KinDiffusion UNet | 12.1 GB | |
|
|
| Both files are required for inference. This repository contains model weights; use them with the official KinDiffusion code. |
|
|
| ## Download |
|
|
| ```bash |
| git clone https://github.com/Unilinear/KinDiffusion.git |
| cd KinDiffusion |
| mkdir -p output/ckpt |
| hf download Unilinear/KinDiffusion \ |
| pytorch_model_3.bin pytorch_model_4.bin \ |
| --local-dir output/ckpt |
| ``` |
|
|
| Then launch the demo: |
|
|
| ```bash |
| python app_kinship.py |
| ``` |
|
|
| See the [project README](https://github.com/Unilinear/KinDiffusion#readme) for environment setup and usage instructions. |
|
|
