media: split the card into three illustrations

- before_after.png: the main one (pose/light/sign text transfer)
- tags_before_after.png: danbooru tags + anime-style prompt
- limitations.png: known issue — object identity (bulldog)

README.md

@@ -33,12 +33,15 @@ Train status: trained on one Vast.ai GPU (RTX 5090, 32 GB), text only — no ima
 [![before/after](media/before_after.png)](media/before_after.png)
 
 Same seed, same settings, same prompt — klein with its native Qwen3-4B encoder on the left,
-klein driven by Qwen3-0.6B + this adapter on the right.
+klein driven by Qwen3-0.6B + this adapter on the right. Pose, light, clothing and the rendered
+sign text all carry over.
+
+## Anime and danbooru tags
 
 [![tags](media/tags_before_after.png)](media/tags_before_after.png)
 
-A danbooru-style tag prompt, same layout. Scene, light, style, pose and rendered text transfer;
-fine identity attributes may drift.
+Both prose and danbooru-style tag prompts work — a tag prompt (top) and an anime-style prompt
+(bottom), native encoder vs. this adapter.
 
 ## What it is
 
@@ -72,7 +75,9 @@ same conditioning into your own pipeline, use `src/klein_condition.py`.
 
 ## Limitations
 
-* Fine identity attributes still drift: a "bulldog" comes out as a terrier-like dog, although the
+[![limitations](media/limitations.png)](media/limitations.png)
+
+* Fine identity attributes still drift: the bulldog above comes out as a terrier-like dog, although
   scene, light, pose and clothing stay faithful. Numerically the adapter keeps improving on these
   prompts (`a bulldog` 29.2% → 26.6% per-token error, `striped bowtie` 29.4% → 25.2%), but the
   breed itself is not transferred yet.
@@ -88,8 +93,9 @@ example.py                   diffusers usage example
 src/adapter_lib.py           adapter schema + loading
 src/klein_condition.py       encode_prompt replacement for the klein pipeline
 src/train_adapter.py         training (online encoders, no cached embeddings)
-media/before_after.png       native Qwen3-4B vs Qwen3-0.6B + adapter
-media/tags_before_after.png  the same for a tag prompt
+media/before_after.png       main illustration: native Qwen3-4B vs Qwen3-0.6B + adapter
+media/tags_before_after.png  danbooru tags and anime-style prompts
+media/limitations.png        known issue: object identity (bulldog)
 ```
 
 ## Donations