Update README.md
Browse files
README.md
CHANGED
|
@@ -43,6 +43,107 @@ hf download jdopensource/JoyAI-Image-Edit-ComfyUI \
|
|
| 43 |
|
| 44 |
Workflow: [workflow_joyimage_edit.json](workflow_joyimage_edit.json)
|
| 45 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 46 |
|
| 47 |
## Links
|
| 48 |
|
|
|
|
| 43 |
|
| 44 |
Workflow: [workflow_joyimage_edit.json](workflow_joyimage_edit.json)
|
| 45 |
|
| 46 |
+
## Spatial Editing Reference
|
| 47 |
+
|
| 48 |
+
JoyAI-Image supports three spatial editing prompt patterns: **Object Move**, **Object Rotation**, and **Camera Control**. For the most stable behavior, we recommend following the prompt templates below as closely as possible.
|
| 49 |
+
For more information (including data curation and evaluation strategies), please refer to [SpatialEdit](https://github.com/EasonXiao-888/SpatialEdit).
|
| 50 |
+
|
| 51 |
+
### Object Move
|
| 52 |
+
|
| 53 |
+
Use this pattern when you want to move a target object into a specified region.
|
| 54 |
+
|
| 55 |
+
**Prompt template:**
|
| 56 |
+
|
| 57 |
+
```text
|
| 58 |
+
Move the <object> into the red box and finally remove the red box.
|
| 59 |
+
```
|
| 60 |
+
|
| 61 |
+
**Rules:**
|
| 62 |
+
|
| 63 |
+
- Replace `<object>` with a clear description of the target object to be moved.
|
| 64 |
+
- The **red box** indicates the target destination in the image.
|
| 65 |
+
- The phrase **"finally remove the red box"** means the guidance box should not appear in the final edited result.
|
| 66 |
+
|
| 67 |
+
**Example:**
|
| 68 |
+
|
| 69 |
+
```text
|
| 70 |
+
Move the apple into the red box and finally remove the red box.
|
| 71 |
+
```
|
| 72 |
+
|
| 73 |
+
### Object Rotation
|
| 74 |
+
|
| 75 |
+
Use this pattern when you want to rotate an object to a specific canonical view.
|
| 76 |
+
|
| 77 |
+
**Prompt template:**
|
| 78 |
+
|
| 79 |
+
```text
|
| 80 |
+
Rotate the <object> to show the <view> side view.
|
| 81 |
+
```
|
| 82 |
+
|
| 83 |
+
**Supported `<view>` values:**
|
| 84 |
+
|
| 85 |
+
```
|
| 86 |
+
front, right, left, rear, front right, front left, rear right, rear left
|
| 87 |
+
```
|
| 88 |
+
|
| 89 |
+
**Rules:**
|
| 90 |
+
|
| 91 |
+
- Replace `<object>` with a clear description of the object to rotate.
|
| 92 |
+
- Replace `<view>` with one of the supported directions above.
|
| 93 |
+
- This instruction is intended to change the **object orientation**, while keeping the object identity and surrounding scene as consistent as possible.
|
| 94 |
+
|
| 95 |
+
**Examples:**
|
| 96 |
+
|
| 97 |
+
```text
|
| 98 |
+
Rotate the chair to show the front side view.
|
| 99 |
+
Rotate the car to show the rear left side view.
|
| 100 |
+
```
|
| 101 |
+
|
| 102 |
+
### Camera Control
|
| 103 |
+
|
| 104 |
+
Use this pattern when you want to change only the camera viewpoint while keeping the 3D scene itself unchanged.
|
| 105 |
+
|
| 106 |
+
**Prompt template:**
|
| 107 |
+
|
| 108 |
+
```text
|
| 109 |
+
Move the camera.
|
| 110 |
+
- Camera rotation: Yaw {y_rotation}°, Pitch {p_rotation}°.
|
| 111 |
+
- Camera zoom: in/out/unchanged.
|
| 112 |
+
- Keep the 3D scene static; only change the viewpoint.
|
| 113 |
+
```
|
| 114 |
+
|
| 115 |
+
**Rules:**
|
| 116 |
+
|
| 117 |
+
- `{y_rotation}` specifies the yaw rotation angle in degrees.
|
| 118 |
+
- `{p_rotation}` specifies the pitch rotation angle in degrees.
|
| 119 |
+
- `Camera zoom` must be one of: `in`, `out`,`unchanged`
|
| 120 |
+
- The last line is important: it explicitly tells the model to preserve the 3D scene content and geometry, and only adjust the camera viewpoint.
|
| 121 |
+
|
| 122 |
+
**Examples:**
|
| 123 |
+
|
| 124 |
+
```text
|
| 125 |
+
Move the camera.
|
| 126 |
+
- Camera rotation: Yaw 45°, Pitch 0°.
|
| 127 |
+
- Camera zoom: in.
|
| 128 |
+
- Keep the 3D scene static; only change the viewpoint.
|
| 129 |
+
```
|
| 130 |
+
|
| 131 |
+
```text
|
| 132 |
+
Move the camera.
|
| 133 |
+
- Camera rotation: Yaw -90°, Pitch 20°.
|
| 134 |
+
- Camera zoom: unchanged.
|
| 135 |
+
- Keep the 3D scene static; only change the viewpoint.
|
| 136 |
+
```
|
| 137 |
+
|
| 138 |
+
### Application
|
| 139 |
+
|
| 140 |
+
**3D Reconstruction:**
|
| 141 |
+
|
| 142 |
+
The first and third examples show point clouds with only a single given viewpoint. The second and fourth examples are augmented by [SpatialEdit](https://github.com/EasonXiao-888/SpatialEdit), which synthesizes richer spatial observations from the sparse input view.
|
| 143 |
+
|
| 144 |
+
**Conditional-frames Based Video Generation:**
|
| 145 |
+
|
| 146 |
+
Given the first frame, [SpatialEdit](https://github.com/EasonXiao-888/SpatialEdit) first generates the final frame of the video, and a video generation model then creates a smooth rotational transition between them while maintaining background consistency.
|
| 147 |
|
| 148 |
## Links
|
| 149 |
|