From 1ac12081c5ade595a005248152ead1aa610178ed Mon Sep 17 00:00:00 2001
From: KATO Kanryu <k.kanryu@gmail.com>
Date: Mon, 25 May 2026 17:04:45 +0900
Subject: [PATCH] Feat: Add documentation for Prompt Token Counter node with
 input details and behavior description

---
 web/docs/TextDrivenPromptTokenCounter.md | 5 +++++
 1 file changed, 5 insertions(+)
 create mode 100644 web/docs/TextDrivenPromptTokenCounter.md

diff --git a/web/docs/TextDrivenPromptTokenCounter.md b/web/docs/TextDrivenPromptTokenCounter.md
new file mode 100644
index 0000000..66b549c
--- /dev/null
+++ b/web/docs/TextDrivenPromptTokenCounter.md
@@ -0,0 +1,5 @@
+### Prompt Token Counter (Text-Driven)
+An input-only visualization node that approximates token counts specifically for Stable Diffusion 1.5 and SDXL workflows.
+* **Inputs:**
+  * `prompt` (String, Force Input): The incoming combined prompt string from upstream nodes.
+* **Behavior:** Utilizing a highly optimized, high-speed estimation algorithm, it dynamically shifts background colors (Green ≤ 65, Yellow 66–75, Red ≥ 76) to visually warn users before prompts breach the critical 75-token CLIP chunk threshold. Tokens forced into the second section or later have a significantly lower probability of being referenced accurately during rendering, potentially leading to image breakdown.