Add fidelity gate note: regeneration often needed after human review
BetterLeaks / scan (push) Successful in 53s
BetterLeaks / scan (push) Successful in 53s
TransitionScene atmosphere match is subjective/fuzzy. Vision extraction is starting point, not guaranteed perfect. Expect 1-2 iterations per courtesan to dial in the feel.
This commit is contained in:
@@ -118,12 +118,16 @@ pausing mid-stride in doorway, half-turned toward viewer with inviting glance
|
||||
2. ✅ Pose reads as "walking transition" (not static standing)
|
||||
3. ✅ Pubic hair color matches head hair
|
||||
4. ✅ Half-turned toward viewer (not full side/back or too frontal)
|
||||
5. ✅ **Subjective atmosphere match:** Does it "feel similar enough" to the Undressed scene? (Fuzzy qualitative gate that can't be perfectly articulated)
|
||||
|
||||
**Reject if:**
|
||||
- ❌ Setting mismatch (wrong room style, modern elements)
|
||||
- ❌ Static pose (looks like Undressed keeper, not transition)
|
||||
- ❌ Wrong hair color
|
||||
- ❌ Pose too side-on or too frontal
|
||||
- ❌ Atmosphere doesn't match (lighting feels off, materials look different, etc.)
|
||||
|
||||
**Important:** Some courtesans may need **regeneration after initial fidelity review** if the extracted setting description didn't capture the right "feel." The vision model extraction is a starting point, not guaranteed perfect match. Expect 1-2 iterations per courtesan to dial in the atmosphere.
|
||||
|
||||
---
|
||||
|
||||
|
||||
Reference in New Issue
Block a user