From 21bdb746ed9939d8829db3280f93e89a67a770e1 Mon Sep 17 00:00:00 2001
From: Olivv <133010426+olivv-cs@users.noreply.github.com>
Date: Thu, 12 Jun 2025 02:41:08 +0300
Subject: [PATCH] Update README.md

---
 README.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/README.md b/README.md
index 81e088fa..640db1fa 100644
--- a/README.md
+++ b/README.md
@@ -22,7 +22,7 @@ Inputs:
 - top_p, top_k, temperature - these are parameters for generating an "assistant prompt";
 - generate_assist_prompt - if disabled, bypasses generation of "assistant prompt", if enable - does it with a model that is load as your text_encoder (might be a custom or a standard one).
 
-Please notice: the encoding runs on CPU because I'm stupid and I have two to three extra minutes to wait for it. You might not have them, but that's you.
+Please notice: after changing a prompt, it's better to call model unloading in Comfy. Believe me.
 Also if you have 24GB of VRAM or less, you might go OOM in Comfy - that's expected, just run the sequence again and it will use cached enhanced prompt.
 
 Technically speaking, it's possible to load just any model as text encoder.