Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,38 @@
|
|
| 1 |
---
|
| 2 |
license: creativeml-openrail-m
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: creativeml-openrail-m
|
| 3 |
---
|
| 4 |
+
|
| 5 |
+
**Prompts:**
|
| 6 |
+
|
| 7 |
+
The model is dreamboothed on tagged suisei no majo images; some prompts that work are
|
| 8 |
+
|
| 9 |
+
1. suletta mercury
|
| 10 |
+
2. miorine rembran
|
| 11 |
+
3. gundam aerial
|
| 12 |
+
|
| 13 |
+
---
|
| 14 |
+
**Training details:**
|
| 15 |
+
|
| 16 |
+
Trained with [kanewallmann Dreambooth repository](https://github.com/kanewallmann/Dreambooth-Stable-Diffusion) using tags as captions
|
| 17 |
+
|
| 18 |
+
1. Trained for 10000 steps probably at the default learning ratet lr=1e-6
|
| 19 |
+
2. Dataset: around 500 tagged images of suise no majo + thousands of customized reg images
|
| 20 |
+
|
| 21 |
+
---
|
| 22 |
+
**Problems:**
|
| 23 |
+
|
| 24 |
+
As the model is trained only on tagged images, it is more flexible but it is only harder to prompt.
|
| 25 |
+
Some detailed description may be needed to get the character right, especially when trying to prompt suletta and miorine in the same image.
|
| 26 |
+
|
| 27 |
+
--
|
| 28 |
+
**Example Generations:**
|
| 29 |
+
|
| 30 |
+
|
| 31 |
+
|
| 32 |
+
|
| 33 |
+
|
| 34 |
+
|
| 35 |
+
|
| 36 |
+
|
| 37 |
+
|
| 38 |
+
|