From 1c1be9460c167c0aa1f5dff95dcdeeb098f6273d Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Jukka=20Sepp=C3=A4nen?=
 <40791699+kijai@users.noreply.github.com>
Date: Mon, 14 Oct 2024 10:53:34 +0300
Subject: [PATCH] Update README.md

---
 README.md | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/README.md b/README.md
index 5d03c97..80c06ca 100644
--- a/README.md
+++ b/README.md
@@ -8,8 +8,8 @@ todo
 - refactor to use comfy text encoding instead
 - optimize memory use
 
-Besides text encoder, this should already run at 10-12GB VRAM when using 1280x768.
-With fp8 and 384p model it can fit under 10GB too.
+Besides text encoder, which can peak at ~12GB VRAM use, this should run at 9-10GB VRAM when using 1280x768.
+With fp8 and 384p model it can fit under 6GB too.
 
 Model loading has not been optimized at all yet, currently needs everything (choose either of the transformers) from here: