Update README.md

README.md

@@ -1,27 +1,65 @@
 ---
 license: apache-2.0
+base_model: stabilityai/stable-diffusion-xl-base-1.0  # ← please add the real base if known
+tags:
+  - lora
+  - 3dcg
+  - anime
+  - realistic
+  - liminal
+  - interior
+  - manga
+  - sfw
+  - nsfw
 ---
-This LoRa is aimed at creating 3DCG female characters in indoor spaces.  With focus on elbow , neck , face , shoulder and hip areas. 
-//--//
-With focus on: Candid movement (face close ups , arm / leg movement , using 3DCG characters from Drills3D and KawaiiDetectiveEnthusiast )
-Interior design (furniture / blur / depth)
-Skin tone (3DCG - suitable for posting on regular SFW as well as NSFW)
-Liminal spaces (Pinterest photos)
-Manga frames (Claymore and Tsutomo Nihei Blame! mixed)
-Photos on adult performers at award events (safe for work! )  by morbidthoughts on flickr. 
-//--//
-This LoRa also helps limitations I have discovered with Chroma. Such as the  'cartoony'  kind of 3D aesthetic ,  in addition to struggles with scifi aesthetic.  I blame the extensive furry training in Chroma. 
-//--//
-Training data are still frames from 3DCG animation into 4x1 sliced 1024x1024 composites.
-//--//
-Ethics: Majority of training data is synthetic content.  This LoRa cannot re-create the likeness of any real world individual.  All photographic trained content are SFW sourced Pinterest photos and Flickr photos.  Training data and selection method are provided in this repository for transparency. 
-//--//
- If you wish to prepare training sets in similiar manner the jupyter colab files have been provided in this repo. 
-//--//
-250 image composites x 4 repeats x 40 epochs
-Rank 32 Alpha 32 at LR= 0.0035 using cosine with repeats
-No bucketing , using Kohya trainer
-//---//
-Use t5xxl_fp16.safetensors text encoder or similiar if using long prompts in Chroma.
-//--//
-This LoRa is hosted on Tensor Art and SeeArt by me (Adcom).  /Cheers!
+
+# 3DCG Indoor Character LoRA
+
+Creates **3DCG-style female characters** in indoor environments with strong emphasis on:
+
+- natural elbow / neck / shoulder posing
+- expressive face close-ups
+- believable hip & torso movement
+- pleasant 3DCG skin rendering (works for both SFW and NSFW)
+
+## Strongest styles / vibes
+
+- Candid 3DCG animation stills (Drills3D / KawaiiDetectiveEnthusiast style)
+- Modern interior design + depth of field / background blur
+- Liminal / backroom / empty mall aesthetic (Pinterest mood)
+- Manga frame composition — Claymore + Tsutomu Nihei (*BLAME!*) influence
+- Red-carpet / award-event posing (SFW photos of adult industry performers)
+
+## Helps fix / compensate for Chroma limitations
+
+- Reduces unwanted cartoony / furry 3D look
+- Improves sci-fi aesthetic compatibility
+- Better anatomy on arms, necks, elbows, hips in longer shots
+
+## Training details
+
+- **Dataset**: ~250 carefully selected 3DCG animation frames  
+  → cropped & composed into 4×1 horizontal 1024×1024 tiles
+- **Repeats × Epochs**: 4 repeats × 40 epochs
+- **Network rank / alpha**: 32 / 32
+- **Learning rate**: 0.0035 · cosine scheduler
+- **Trainer**: Kohya_ss (no bucketing)
+- Mostly synthetic data + SFW Pinterest / Flickr reference photos
+
+## Ethics & transparency
+
+- Cannot reproduce real people (no real-person likenesses)
+- All photographic content used = public SFW images only
+- Training image selection & preparation scripts → included in this repo (Jupyter colab notebooks)
+
+## Recommended usage in Chroma
+
+```text
+Trigger word(s): 3dcg-girl, 3dcg-style  (optional but helps)
+
+Best text encoder: t5xxl_fp16.safetensors   (especially with long descriptive prompts)
+
+Weight range:   0.65 – 0.95   (0.75–0.85 usually sweetest spot)
+
+Negative prompt pieces that often help:
+low quality, furry, anthro, cartoon, 2d, flat color, sketch