File size: 2,101 Bytes
d548f90 2ef6ae4 d548f90 2492234 d548f90 2ef6ae4 efc26de 6ec1beb d548f90 2ef6ae4 d548f90 2ef6ae4 d548f90 2ef6ae4 d548f90 2ef6ae4 d548f90 71d1fe8 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 |
# SD3 Controlnet
| control image | weight=0.0 | weight=0.3 | weight=0.5 | weight=0.7 | weight=0.9 |
|:-------------------------:|:-------------------------:|:-------------------------:|:-------------------------:|:-------------------------:|:-------------------------:|
|<img src="./pose.jpg" width = "400" /> | <img src="./demo_0.jpg" width = "400" /> | <img src="./demo_3.jpg" width = "400" /> | <img src="./demo_5.jpg" width = "400" /> | <img src="./demo_7.jpg" width = "400" /> | <img src="./demo_9.jpg" width = "400" /> |
# Install Diffusers-SD3-Controlnet
The current [diffusers](https://github.com/instantX-research/diffusers_sd3_control.git) have not been merged into the official code yet.
```cmd
git clone -b sd3_control https://github.com/instantX-research/diffusers_sd3_control.git
cd diffusers_sd3_control
pip install -e .
```
# Demo
```python
import torch
from diffusers import StableDiffusion3ControlNetPipeline
from diffusers.models import SD3ControlNetModel, SD3MultiControlNetModel
from diffusers.utils import load_image
# load pipeline
controlnet = SD3ControlNetModel.from_pretrained("InstantX/SD3-Controlnet-Pose")
pipe = StableDiffusion3ControlNetPipeline.from_pretrained(
"stabilityai/stable-diffusion-3-medium-diffusers",
controlnet=controlnet
)
pipe.to("cuda", torch.float16)
# config
control_image = load_image("https://huggingface.co/InstantX/SD3-Controlnet-Pose/resolve/main/pose.jpg")
prompt = 'Anime style illustration of a girl wearing a suit. A moon in sky. In the background we see a big rain approaching. text "InstantX" on image'
n_prompt = 'NSFW, nude, naked, porn, ugly'
image = pipe(
prompt,
negative_prompt=n_prompt,
control_image=pose_image,
controlnet_conditioning_scale=0.5,
).images[0]
image.save('image.jpg')
```
## Limitation
Due to the fact that only 1024*1024 pixel resolution was used during the training phase,
the inference performs best at this size, with other sizes yielding suboptimal results.
We will initiate multi-resolution training in the future, and at that time, we will open-source the new weights.
|