---
library_name: peft
tags:
  - image-editing
  - lora
  - robotics
---

# HandEdit LoRA

LoRA checkpoints for replacing human hands with Inspire robot hands while
preserving the original pose, object interaction, background, lighting, and
camera view.

The adapters were trained on 20K+ aligned image-editing pairs.

## Models

| Model | Base model | Checkpoint |
| --- | --- | --- |
| LongCat-Image-Edit | `meituan-longcat/LongCat-Image-Edit` | `checkpoints/longcat/` |
| OmniGen-v1 | `Shitao/OmniGen-v1` | `checkpoints/omnigen/` |
| FLUX.2 Klein Base 4B | `black-forest-labs/FLUX.2-klein-base-4B` | `checkpoints/flux2/` |
| Step1X-Edit | `stepfun-ai/Step1X-Edit` | `checkpoints/step1x/` |

## Download

```bash
hf download HandEdit/HandEdit-LoRA \
  --include "checkpoints/**" \
  --local-dir .
```

## Inference

Run from the repository root:

```bash
python scripts/infer_longcat_lora.py \
  --input_dir ./input --output_dir ./output/longcat

python scripts/infer_omnigen_lora.py \
  --input_dir ./input --output_dir ./output/omnigen

python scripts/infer_flux2_lora.py \
  --input_dir ./input --output_dir ./output/flux2

python scripts/infer_step1x_lora.py \
  --repo_dir ./third_party/Step1X-Edit \
  --model_dir ./weights/Step1X-Edit \
  --input_dir ./input --output_dir ./output/step1x \
  --quantized --offload
```

Only the LoRA adapters are included. Download the corresponding base models
from their official repositories.