From 6de73fcedcccd64843a5f615e1bc7bdb7c8238b3 Mon Sep 17 00:00:00 2001 From: Oliver Switzer Date: Mon, 11 Aug 2025 09:11:03 -0400 Subject: [PATCH] docs: enhance IMAGE_PROMPT guidelines in PromptSplitterNode --- nodes/prompt_splitter_node.py | 10 ++++++++++ 1 file changed, 10 insertions(+) diff --git a/nodes/prompt_splitter_node.py b/nodes/prompt_splitter_node.py index f4dd20b..5fd4c61 100644 --- a/nodes/prompt_splitter_node.py +++ b/nodes/prompt_splitter_node.py @@ -76,6 +76,16 @@ class PromptSplitterNode: _SYSTEM_PROMPT = """ You are splitting a single combined scene description into two separate outputs: +💡 Best Practices for IMAGE_PROMPT (SD image generation): +- Use a keyword/CSV-style format, separated by commas. +- Keep descriptors short, concrete, and visual — e.g., "woman, 4k, red dress, rooftop at night, string lights". +- Preserve ordering so the main subject comes first, followed by descriptors, environment, style cues. +- Avoid writing full sentences or sequential events — this is a still image, not a scene description. +- Keep detail density high (pose, clothing, props, lighting) but ONLY from source text. +- Do NOT insert adjectives or style tags unless they already appear in the input. + +--- + 1. IMAGE_PROMPT — For SD image generation. - Keep only static, visual descriptors: characters, clothing, environment, props, positions, facial expressions (if static). - Do NOT add or remove descriptive terms unless they already appear in the source.