Upload 2368299/2663428/README.md with huggingface_hub
Browse files- 2368299/2663428/README.md +104 -3
2368299/2663428/README.md
CHANGED
|
@@ -1,10 +1,111 @@
|
|
| 1 |
---
|
| 2 |
license: other
|
| 3 |
tags:
|
| 4 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 5 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
Author: [BitcrushedHeart](https://civitai.com/user/BitcrushedHeart)
|
| 7 |
|
| 8 |
-
Model: [
|
|
|
|
|
|
|
| 9 |
|
| 10 |
-
|
|
|
|
| 1 |
---
|
| 2 |
license: other
|
| 3 |
tags:
|
| 4 |
+
- amateur
|
| 5 |
+
- candid
|
| 6 |
+
- concept
|
| 7 |
+
- natural body
|
| 8 |
+
- nsfw
|
| 9 |
+
- nudity
|
| 10 |
---
|
| 11 |
+
|
| 12 |
+
# SoReal! Natural Bodies - v1.0 - 2663428
|
| 13 |
+
|
| 14 |
+
**Model Type**: LORA
|
| 15 |
+
|
| 16 |
+
**Base Model**: ZImageBase
|
| 17 |
+
|
| 18 |
+
**Trigger Words**: None
|
| 19 |
+
|
| 20 |
+
**Tags**: amateur, candid, concept, natural body, nsfw, nudity
|
| 21 |
+
|
| 22 |
+
## Gallery
|
| 23 |
+
|
| 24 |
+
<table>
|
| 25 |
+
<tr>
|
| 26 |
+
<td><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/4099fa69-5df2-4b3e-9da4-3fd55560cc30/original=true/120065743.jpeg" width="200"></td>
|
| 27 |
+
<td><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/a9b20d95-d4c1-4f00-ad39-f3b4f11c40ce/original=true/120065787.jpeg" width="200"></td>
|
| 28 |
+
<td><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/7712b15a-dcac-4bfd-a3b6-53fbb592323f/original=true/120065803.jpeg" width="200"></td>
|
| 29 |
+
</tr>
|
| 30 |
+
<tr>
|
| 31 |
+
<td><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/b113b8d3-0b5d-47e0-8d8d-3dff61f0fb6a/original=true/120065818.jpeg" width="200"></td>
|
| 32 |
+
<td><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/43688524-ab40-45df-ae25-79a3572d75da/original=true/120065844.jpeg" width="200"></td>
|
| 33 |
+
<td><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/2f878835-ad96-489b-af8d-759d74f1b86c/original=true/120065890.jpeg" width="200"></td>
|
| 34 |
+
</tr>
|
| 35 |
+
<tr>
|
| 36 |
+
<td><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/d09c14dd-0d0c-4bb3-b308-64f4b4fa683d/original=true/120065920.jpeg" width="200"></td>
|
| 37 |
+
<td><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/d7415523-3cb4-40f4-84b5-709c2ef45654/original=true/120065964.jpeg" width="200"></td>
|
| 38 |
+
<td><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/f0e12399-f6e4-48b9-9cce-20020c409e8c/original=true/120066032.jpeg" width="200"></td>
|
| 39 |
+
</tr>
|
| 40 |
+
<tr>
|
| 41 |
+
<td><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/58d7eee3-aab5-4e43-b209-0871466cd342/original=true/120066065.jpeg" width="200"></td>
|
| 42 |
+
</tr>
|
| 43 |
+
</table>
|
| 44 |
+
|
| 45 |
+
## Description
|
| 46 |
+
|
| 47 |
+
[**Follow me on Patreon!**](https://www.patreon.com/Bitcrushed_Heart)
|
| 48 |
+
---------------------------------------------------------------------
|
| 49 |
+
|
| 50 |
+
**SoReal! - Natural Bodies**
|
| 51 |
+
============================
|
| 52 |
+
|
| 53 |
+
[[SoReal! Portraits]](https://civitai.com/models/2360068) [[SoReal POV]](https://civitai.com/models/2357375?modelVersionId=2657257)
|
| 54 |
+
|
| 55 |
+
**Overview**
|
| 56 |
+
------------
|
| 57 |
+
|
| 58 |
+
Stand aside, supermodels! This model is the next iteration of my Z-Image LORAs - aimed to bring diversity in both concepts and humanity itself to Z-Image.
|
| 59 |
+
|
| 60 |
+
**Compatibility & Usage**
|
| 61 |
+
-------------------------
|
| 62 |
+
|
| 63 |
+
Due to it's small size and rank, the model should have a minimal influence on the base model, further improving compatibility with other LORAs across Base/Turbo and indeed other checkpoints.
|
| 64 |
+
|
| 65 |
+
'Trigger words' aren't real - don't ask for one, just prompt normally! Ages (xx-year-old format), weight classifications (underweight, natural weight, average weight, obese, etc), ethnicity and skin tones were used in captioning based on real data (rather than estimations).
|
| 66 |
+
|
| 67 |
+
When using Z-Image Turbo, strengths between 0.90 and 1.2 seem to have good results.
|
| 68 |
+
|
| 69 |
+
**Limitations**
|
| 70 |
+
---------------
|
| 71 |
+
|
| 72 |
+
TBC
|
| 73 |
+
|
| 74 |
+
**Future**
|
| 75 |
+
----------
|
| 76 |
+
|
| 77 |
+
**I am planning on finetuning Z-Image considerably with a model called 'SoReal!' (Or, alternatively, ZoReal!). However, I want it to be the best possible amateur finetune possible, to achieve this, I have:**
|
| 78 |
+
|
| 79 |
+
* 1. Trained a custom quality model.
|
| 80 |
+
* 2. Trained a custom one-shot demographic model (height, weight, skin tone, ethnicity, age in years, body shape) with an average accuracy of 89% for top-confidence prediction using ConvNext-XL.
|
| 81 |
+
* 3. Finetuned wd-tagger-large-v3 on a large sample dataset of 50k hand-tagged images with human-assisted active learning.
|
| 82 |
+
* 4. Fed those tagged images (with quality, demographics and general labels) with the image metadata (incl. EXIF & Camera Metadata) to Gemini 3 Flash for generating captions.
|
| 83 |
+
* No over-trained LORAs baked in, no dramatic loss of generalisation, just a good, all-round, NSFW-ready, finetuned model.
|
| 84 |
+
|
| 85 |
+
**I am now severely limited, however, by my compute and financial situation, so if you'd like to help make SoReal!, well, so real,** [**then you can follow me on Patreon!**](https://www.patreon.com/Bitcrushed_Heart)
|
| 86 |
+
|
| 87 |
+
**Dataset & Training**
|
| 88 |
+
----------------------
|
| 89 |
+
|
| 90 |
+
Dataset of 1000 sourced from a variety of sources. Deduplication and Quality Scoring (through MANIQA) batch size 16.
|
| 91 |
+
|
| 92 |
+
Validation loss was used with 10% of the dataset size to prevent overfitting while still maintaining strong concept adherence and generalisation.
|
| 93 |
+
|
| 94 |
+
Model was trained with AdamW through the Python adv-optm package.
|
| 95 |
+
|
| 96 |
+
**Licensing**
|
| 97 |
+
-------------
|
| 98 |
+
|
| 99 |
+
If you'd like to release a merge of this model, please contact me.
|
| 100 |
+
|
| 101 |
+
Made with <3 By BitcrushedHeart
|
| 102 |
+
|
| 103 |
+
---
|
| 104 |
+
|
| 105 |
Author: [BitcrushedHeart](https://civitai.com/user/BitcrushedHeart)
|
| 106 |
|
| 107 |
+
Model: [CivitAI Model Page](https://civitai.com/models/2368299?modelVersionId=2663428)
|
| 108 |
+
|
| 109 |
+
Archive: [CivArchive Page](https://civarchive.com/models/2368299?modelVersionId=2663428)
|
| 110 |
|
| 111 |
+
<!-- Version: 20260422_v4_update -->
|