Refine visual identity guidance across pipelines
This commit is contained in:
@@ -102,9 +102,11 @@ When mixing stock and generated images in the same video, visual consistency is
|
||||
Apply the playbook's color grading LUT to both stock and generated images in the compose stage.
|
||||
This unifies the look. The `color_grade` enhancement tool handles this.
|
||||
|
||||
### Strategy: Match Playbook Style in Prompts
|
||||
When generating images, always prepend the playbook's `image_prompt_prefix`. When searching stock,
|
||||
use the playbook's color names in search filters (Pexels supports color filtering).
|
||||
### Strategy: Match The Visual Identity, Not A Copied Prefix
|
||||
When generating images, adapt the playbook's mood, palette, texture, and medium
|
||||
into a short scene-specific anchor. When searching stock, filter for the same
|
||||
emotional and visual qualities: color, lighting, environment, composition, era,
|
||||
and texture. Use the playbook as a consistency source, not a script to paste.
|
||||
|
||||
### Strategy: Avoid Mixing Styles Within a Scene
|
||||
Don't use a stock photo for one element and an AI illustration for another in the same scene.
|
||||
|
||||
Reference in New Issue
Block a user