feature: pixel party xl v1.0 model release
Browse files- README.md +65 -0
- config.json +54 -0
- diffusion_pytorch_model.safetensors +3 -0
- images/examples.gif +0 -0
README.md
ADDED
@@ -0,0 +1,65 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: creativeml-openrail-m
|
3 |
+
tags:
|
4 |
+
- text-to-image
|
5 |
+
- stable-diffusion
|
6 |
+
- diffusers
|
7 |
+
base_model: stabilityai/stable-diffusion-xl-base-1.0
|
8 |
+
instance_prompt: . in pixel art style
|
9 |
+
widget:
|
10 |
+
- text: cute dragon. in pixel art style
|
11 |
+
---
|
12 |
+
|
13 |
+
# Pixel Party XL
|
14 |
+
|
15 |
+
This is a full model training for better pixel art adherence based on SDXL. Feel free to use this model for your own projects, but please do not host it.
|
16 |
+
|
17 |
+
![examples](images/examples.gif)
|
18 |
+
|
19 |
+
We are building on tools for indie game development and currently have tools:
|
20 |
+
|
21 |
+
- Map tiles
|
22 |
+
- Movement animations
|
23 |
+
- Attack animations
|
24 |
+
- Inpainting
|
25 |
+
- Character reshaping
|
26 |
+
- Animation interpolation
|
27 |
+
|
28 |
+
And have much more planned! :D
|
29 |
+
|
30 |
+
If you want to support us or check out our other pixel art models, you can find us here [PixelLab](https://www.pixellab.ai) or on [Discord](https://discord.gg/pBeyTBF8T7).
|
31 |
+
|
32 |
+
## How to use
|
33 |
+
|
34 |
+
- Append ". in pixel art style" to your prompt. E.g. "cute dragon. in pixel art style"
|
35 |
+
- Downsize the image 8x using nearest neighbor
|
36 |
+
- Init images are very helpful
|
37 |
+
- Model works best at around 128x128 canvas size but still excels at creating smaller items/characters/other
|
38 |
+
- Use a VAE with fixed fp16 support: https://huggingface.co/madebyollin/sdxl-vae-fp16-fix
|
39 |
+
- Do not use refiner
|
40 |
+
|
41 |
+
### Diffusers
|
42 |
+
|
43 |
+
```python
|
44 |
+
from diffusers import DiffusionPipeline, UNet2DConditionModel
|
45 |
+
import torch
|
46 |
+
|
47 |
+
pipe = DiffusionPipeline.from_pretrained(
|
48 |
+
"stabilityai/stable-diffusion-xl-base-1.0",
|
49 |
+
torch_dtype=torch.float16,
|
50 |
+
unet=UNet2DConditionModel.from_pretrained("pixel-party-v1.0", torch_dtype=torch.float16),
|
51 |
+
use_safetensors=True,
|
52 |
+
variant="fp16",
|
53 |
+
)
|
54 |
+
pipe.to("cuda")
|
55 |
+
|
56 |
+
torch.manual_seed(11215)
|
57 |
+
prompt = "cute dragon. in pixel art style"
|
58 |
+
negative_prompt = "mixels. amateur. multiple"
|
59 |
+
|
60 |
+
image = pipe(prompt, negative_prompt=negative_prompt, num_inference_steps=25).images[0]
|
61 |
+
```
|
62 |
+
|
63 |
+
## License
|
64 |
+
|
65 |
+
Please do not host this model. It is otherwise licensed under CreativeML-OpenRail-M.
|
config.json
ADDED
@@ -0,0 +1,54 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_class_name": "UNet2DConditionModel",
|
3 |
+
"_diffusers_version": "0.19.3",
|
4 |
+
"_name_or_path": "pixel-party-xl-v1.0/denoiser/unet",
|
5 |
+
"act_fn": "silu",
|
6 |
+
"addition_embed_type": "text_time",
|
7 |
+
"addition_embed_type_num_heads": 64,
|
8 |
+
"addition_time_embed_dim": 256,
|
9 |
+
"attention_head_dim": [5, 10, 20],
|
10 |
+
"block_out_channels": [320, 640, 1280],
|
11 |
+
"center_input_sample": false,
|
12 |
+
"class_embed_type": null,
|
13 |
+
"class_embeddings_concat": false,
|
14 |
+
"conv_in_kernel": 3,
|
15 |
+
"conv_out_kernel": 3,
|
16 |
+
"cross_attention_dim": 2048,
|
17 |
+
"cross_attention_norm": null,
|
18 |
+
"down_block_types": [
|
19 |
+
"DownBlock2D",
|
20 |
+
"CrossAttnDownBlock2D",
|
21 |
+
"CrossAttnDownBlock2D"
|
22 |
+
],
|
23 |
+
"downsample_padding": 1,
|
24 |
+
"dual_cross_attention": false,
|
25 |
+
"encoder_hid_dim": null,
|
26 |
+
"encoder_hid_dim_type": null,
|
27 |
+
"flip_sin_to_cos": true,
|
28 |
+
"freq_shift": 0,
|
29 |
+
"in_channels": 4,
|
30 |
+
"layers_per_block": 2,
|
31 |
+
"mid_block_only_cross_attention": null,
|
32 |
+
"mid_block_scale_factor": 1,
|
33 |
+
"mid_block_type": "UNetMidBlock2DCrossAttn",
|
34 |
+
"norm_eps": 1e-5,
|
35 |
+
"norm_num_groups": 32,
|
36 |
+
"num_attention_heads": null,
|
37 |
+
"num_class_embeds": null,
|
38 |
+
"only_cross_attention": false,
|
39 |
+
"out_channels": 4,
|
40 |
+
"projection_class_embeddings_input_dim": 2816,
|
41 |
+
"resnet_out_scale_factor": 1.0,
|
42 |
+
"resnet_skip_time_act": false,
|
43 |
+
"resnet_time_scale_shift": "default",
|
44 |
+
"sample_size": 128,
|
45 |
+
"time_cond_proj_dim": null,
|
46 |
+
"time_embedding_act_fn": null,
|
47 |
+
"time_embedding_dim": null,
|
48 |
+
"time_embedding_type": "positional",
|
49 |
+
"timestep_post_act": null,
|
50 |
+
"transformer_layers_per_block": [1, 2, 10],
|
51 |
+
"up_block_types": ["CrossAttnUpBlock2D", "CrossAttnUpBlock2D", "UpBlock2D"],
|
52 |
+
"upcast_attention": null,
|
53 |
+
"use_linear_projection": true
|
54 |
+
}
|
diffusion_pytorch_model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ada8d88a0a7fe5417d4d63ee4a66897383ec682c1cd54f083acbfd47a57e7779
|
3 |
+
size 5135149760
|
images/examples.gif
ADDED