Fix: flatten comfyui and ltx-video-pipeline to Gitea convention (flat skill dirs). comfyui v5.1.0, ltx-video-pipeline v2.0.0.
This commit is contained in:
@@ -0,0 +1,24 @@
|
||||
# Canonical Character Descriptions
|
||||
|
||||
Ground truth for all LTX prompts and Flux panorama generation. Do NOT improvise or use generic descriptions — these are the exact descriptions the user provided.
|
||||
|
||||
## Boss
|
||||
|
||||
Black suit, muscular build, shaved head, goatee. Stern expression, chrome temple implant. Stock reference: `SHEETS2_00005_Boss.png` on TrueNAS.
|
||||
|
||||
**Prompt fragment:** "A muscular boss with shaved head and goatee wearing a black suit sits at a terminal in a neon-lit cyberpunk room. His chrome temple implant glints in the neon glow. Stern expression."
|
||||
|
||||
## Woman
|
||||
|
||||
Mid-20s, short dark hair, chrome temple implants, cybernetic fingers, black synth-leather jacket. Stock references: `cyberpunk_woman_neon_01.jpg`, `cyberpunk_woman_neon_02.jpg` on TrueNAS.
|
||||
|
||||
**Prompt fragment:** "A woman in her mid-20s with short dark hair, chrome temple implants, and cybernetic fingers wears a black synth-leather jacket. She stands by a window gazing out at the neon cityscape."
|
||||
|
||||
## Pitfall: Generic Descriptions
|
||||
|
||||
Using generic descriptions like "stern boss" or "woman by window" instead of the canonical descriptions causes:
|
||||
- Face distortion (model doesn't know what "stern boss" looks like)
|
||||
- Character morphing (model blends between prompt and start frame)
|
||||
- Wrong character entirely (model substitutes its own training data)
|
||||
|
||||
Always use the exact canonical descriptions above in every prompt that references these characters.
|
||||
Reference in New Issue
Block a user