From 7b384a0643cec1cae1758fb0d2e066e03fd39106 Mon Sep 17 00:00:00 2001
From: digital-garbage <133010426+digital-garbage@users.noreply.github.com>
Date: Mon, 18 May 2026 20:42:01 +0300
Subject: [PATCH] Creativity mask applies at all reward levels including
 Perfect

Removed reward >= 0.8 early returns from build_creativity_mask and
load_and_apply_creativity_mask. The global floor formula already handles
high reward correctly: at 1.0 it produces 0.0 floor, leaving only gentle
spatial variance (0-8% of latent std). A Perfect-rated latent is the best
possible reference - discarding it was wrong.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---
 ltx_enhancements.py | 12 ++++--------
 1 file changed, 4 insertions(+), 8 deletions(-)

diff --git a/ltx_enhancements.py b/ltx_enhancements.py
index ea9eee74..7d0a0c7a 100644
--- a/ltx_enhancements.py
+++ b/ltx_enhancements.py
@@ -92,10 +92,8 @@ def load_and_apply_creativity_mask(refinement_key, rating_profile, reward, laten
     """
     Computes creativity mask and returns noise-modified latent dict or None.
     Source priority: connected latent arg > saved file by key > None.
-    Returns None when reward >= 0.8 (no need to push harder when things work).
+    At high reward the global floor is 0 and only gentle spatial variance applies.
     """
-    if reward >= 0.8:
-        return None
 
     # Resolve source latent
     samples = None
@@ -225,12 +223,10 @@ def _derive_temperature_map(rating_profile, reward):
 
 def build_creativity_mask(latent, rating_profile, reward):
     """
-    Returns a denoise_mask tensor [B, T, H, W] or None if no adjustment needed.
-    High-variance latent regions get more creative freedom (higher mask values).
-    Rating drives global floor: worse rating = more freedom everywhere.
+    Returns a noise scale mask or None if latent is unusable.
+    High-variance regions get more creative freedom. Global floor is 0 at
+    high reward (formula clamps naturally) and rises toward 0.35 at reward -1.0.
     """
-    if reward >= 0.8:
-        return None
 
     samples = latent.get("samples") if isinstance(latent, dict) else None
     if not isinstance(samples, torch.Tensor):