From b7e1b5ad79b7fc18e727412e3829aa075f4ecc97 Mon Sep 17 00:00:00 2001
From: digital-garbage <133010426+digital-garbage@users.noreply.github.com>
Date: Sun, 23 Aug 2026 08:39:10 +0300
Subject: [PATCH] =?UTF-8?q?docs(h3):=20the=20phrase=20emphasis=20has=20no?=
 =?UTF-8?q?=20toggle=20=E2=80=94=20say=20what=20actually=20gates=20it?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 conditioning.py | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/conditioning.py b/conditioning.py
index 33aea36a..03cf4898 100644
--- a/conditioning.py
+++ b/conditioning.py
@@ -13767,8 +13767,11 @@ class FunPackVideoRefinerV2(FunPackVideoRefiner):
 
         This writes it as (start, end, weight) token spans for the sampler to turn into an
         attention-logit bias. Text-only and cheap: phrases are located with the tokenizer's
-        offset mapping, not by re-encoding each one through a 32B text encoder. No behaviour
-        change unless the sampler's toggle is on.
+        offset mapping, not by re-encoding each one through a 32B text encoder.
+
+        There is no toggle. It is gated on having something to say: an H3 CLIP, a phrase
+        memory, and a rating that named a missing axis. Before the first rating it produces
+        nothing, and a Perfect clears the axes and switches it off again.
         """
         if clip is None or not isinstance(phrase_memory, dict) or not phrase_memory:
             return conditioning_list