Update README.md
Browse files
README.md
CHANGED
|
@@ -32,11 +32,34 @@ license: mit
|
|
| 32 |
<Gallery />
|
| 33 |
|
| 34 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 35 |
## Trigger words
|
| 36 |
|
| 37 |
You should use `an anime illustration of` to trigger the image generation.
|
| 38 |
|
| 39 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 40 |
## Download model
|
| 41 |
|
| 42 |
Weights for this model are available in Safetensors format.
|
|
|
|
| 32 |
<Gallery />
|
| 33 |
|
| 34 |
|
| 35 |
+
A experimental model that uses HunyuanVideo as a image generator. outputs images at 768 resolution.
|
| 36 |
+
|
| 37 |
+
In a typical HunyuanVideo workflow, set 'frame' to **1** and add this lora to get an anime illustration-like output.
|
| 38 |
+
|
| 39 |
+
|
| 40 |
## Trigger words
|
| 41 |
|
| 42 |
You should use `an anime illustration of` to trigger the image generation.
|
| 43 |
|
| 44 |
|
| 45 |
+
## Resolutions
|
| 46 |
+
|
| 47 |
+
Use the following resolution for the best results:
|
| 48 |
+
```
|
| 49 |
+
(768, 768)
|
| 50 |
+
(672, 864), (864, 672)
|
| 51 |
+
(608, 960), (960, 608)
|
| 52 |
+
(544, 1088), (1088, 544)
|
| 53 |
+
```
|
| 54 |
+
|
| 55 |
+
## Training
|
| 56 |
+
|
| 57 |
+
The model has been trained on a tag-balanced dataset of 2k best pixiv illustrations, at resolution of 768, for 856 eopchs. (214e * repeat4)
|
| 58 |
+
|
| 59 |
+
The training takes about 4 days on a 8xH100 cluster.
|
| 60 |
+
|
| 61 |
+
|
| 62 |
+
|
| 63 |
## Download model
|
| 64 |
|
| 65 |
Weights for this model are available in Safetensors format.
|