Update README.md
Browse files
README.md
CHANGED
|
@@ -43,3 +43,114 @@ Use case examples:
|
|
| 43 |
jackie_0010_00001_.png
|
| 44 |
jackie_0010_00001_.txt
|
| 45 |
...
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 43 |
jackie_0010_00001_.png
|
| 44 |
jackie_0010_00001_.txt
|
| 45 |
...
|
| 46 |
+
Repeats: 20
|
| 47 |
+
|
| 48 |
+
Effective train images with repeats: 540
|
| 49 |
+
|
| 50 |
+
Hyperparameters (key ones):
|
| 51 |
+
|
| 52 |
+
Resolution: 512x512, buckets enabled, bucket_no_upscale = true
|
| 53 |
+
|
| 54 |
+
Epochs: 3
|
| 55 |
+
|
| 56 |
+
Max train steps: 800
|
| 57 |
+
|
| 58 |
+
Batch size: 2
|
| 59 |
+
|
| 60 |
+
Optimizer: AdamW8bit
|
| 61 |
+
|
| 62 |
+
Learning rate: 5e-4
|
| 63 |
+
|
| 64 |
+
LR scheduler: cosine
|
| 65 |
+
|
| 66 |
+
Loss type: l2
|
| 67 |
+
|
| 68 |
+
Mixed precision: fp16
|
| 69 |
+
|
| 70 |
+
Xformers: enabled
|
| 71 |
+
|
| 72 |
+
No regularization images used
|
| 73 |
+
|
| 74 |
+
Checkpoints:
|
| 75 |
+
|
| 76 |
+
jackieV1-000001.safetensors (epoch 1)
|
| 77 |
+
|
| 78 |
+
jackieV1-000002.safetensors (epoch 2)
|
| 79 |
+
|
| 80 |
+
jackieV1.safetensors (final embedding)
|
| 81 |
+
|
| 82 |
+
How to Use
|
| 83 |
+
Automatic1111
|
| 84 |
+
Download jackieV1.safetensors.
|
| 85 |
+
|
| 86 |
+
Place it into:
|
| 87 |
+
|
| 88 |
+
text
|
| 89 |
+
Copy code
|
| 90 |
+
stable-diffusion-webui/embeddings/
|
| 91 |
+
Restart the UI.
|
| 92 |
+
|
| 93 |
+
Use the trigger token jck1 in your prompt.
|
| 94 |
+
|
| 95 |
+
Example prompt (portrait):
|
| 96 |
+
|
| 97 |
+
text
|
| 98 |
+
Copy code
|
| 99 |
+
jck1, woman, portrait, upper body, neutral expression,
|
| 100 |
+
soft studio lighting, detailed skin, sharp focus, 8k
|
| 101 |
+
Example negative prompt:
|
| 102 |
+
|
| 103 |
+
text
|
| 104 |
+
Copy code
|
| 105 |
+
low quality, bad anatomy, extra limbs, distorted face, blurry, grainy
|
| 106 |
+
Example prompt (full body, scene):
|
| 107 |
+
|
| 108 |
+
text
|
| 109 |
+
Copy code
|
| 110 |
+
jck1, woman, full body, casual clothing, standing on a city sidewalk at sunset,
|
| 111 |
+
cinematic lighting, shallow depth of field, realistic, detailed skin, 8k
|
| 112 |
+
ComfyUI
|
| 113 |
+
Copy jackieV1.safetensors to:
|
| 114 |
+
|
| 115 |
+
text
|
| 116 |
+
Copy code
|
| 117 |
+
ComfyUI/models/embeddings/jackieV1.safetensors
|
| 118 |
+
Restart ComfyUI so it rescans models.
|
| 119 |
+
|
| 120 |
+
In your CLIP text node, include jck1 in the positive prompt.
|
| 121 |
+
|
| 122 |
+
Example ComfyUI positive prompt:
|
| 123 |
+
|
| 124 |
+
text
|
| 125 |
+
Copy code
|
| 126 |
+
jck1, woman, portrait, upper body, soft studio lighting,
|
| 127 |
+
cinematic, ultra detailed, SDXL, juggernaut style
|
| 128 |
+
Pair with your usual negative prompt / sampler / CFG settings.
|
| 129 |
+
|
| 130 |
+
Suggested Settings
|
| 131 |
+
These are starting points only; tweak to your taste.
|
| 132 |
+
|
| 133 |
+
Base model: juggernautXL_ragnarokBy.safetensors (SDXL)
|
| 134 |
+
|
| 135 |
+
Sampler: Euler a / DPM++ 2M Karras
|
| 136 |
+
|
| 137 |
+
Steps: 25–35
|
| 138 |
+
|
| 139 |
+
CFG scale: 5–8
|
| 140 |
+
|
| 141 |
+
Resolution: 768×1024 or 832×1216 (portrait) / similar area for landscape
|
| 142 |
+
|
| 143 |
+
For strong identity lock, keep jck1 toward the front of the prompt and
|
| 144 |
+
avoid stacking too many style tokens that overpower it.
|
| 145 |
+
|
| 146 |
+
Notes, Rights, and Usage
|
| 147 |
+
This embedding was trained on a private Jackie dataset.
|
| 148 |
+
|
| 149 |
+
You are responsible for ensuring that your use of this embedding respects the
|
| 150 |
+
rights and privacy of any real person whose likeness it may represent.
|
| 151 |
+
|
| 152 |
+
Do not use this model for harassment, defamation, deepfake abuse, or any
|
| 153 |
+
malicious activity.
|
| 154 |
+
|
| 155 |
+
If you create cool results or derivative work from this embedding, feel free to
|
| 156 |
+
tag @BonusLockSmith.
|