Universal Studio Prompt Director (Master Encyclopedic Hub)

The UniversalStudioPromptDirector is the pack's "master hub" - the node that chains four of the other five together so a whole character-scene workflow hangs off one box. In one node it casts two characters, writes a multi-reference scene prompt, writes an LTX video prompt with dialogue and SFX, and produces a negative prompt. Five strings out, ready to wire into your image and video pipelines: char1_casting_prompt, char2_casting_prompt, multi_ref_scene_prompt, ltx_video_prompt, negative_prompt.

If you're a beginner this is the most approachable entry point in the pack, because it comes pre-loaded with a working scene. The defaults are Maya (3D Pixar young woman, bob hair, beige tee) and Mariam (26-year-old Egyptian woman in a pastel pink hijab, photoreal) having a heated coffee-shop argument that escalates to a slap. It's a fully coherent prompt set out of the box - load the two reference images, drop the outputs into your Klein and LTX workflows, and you have a test render before you've changed anything.

How it works

Each of the two character presets (character_1_preset, character_2_preset) expands into a full casting-style description - the photoreal ones read like an actor's brief, with skin texture, eye color, lens and lighting baked in. That's genuinely useful for the character-consistency problem: the same casting prompt reused across shots keeps a face far more stable than ad-libbing a description every time. scene_location_preset and scene_drama_climax (argument climax, verbal confrontation, casual conversation, tearful reconciliation) feed the scene and video prompts. The negative_prompt_preset reuses the same guard set as the Flux builder.

The honest caveat: ltx_video_prompt is opinionated in a way that will surprise you the first time. It's hard-wired around the argument/slap scenario - Maya slaps Mariam, Mariam recoils clutching her cheek - regardless of which drama preset you pick. You customize it through the Dialogue_Script_English and Custom_Director_Notes boxes, not through a choreography dropdown. That's fine when the scene fits; just don't expect it to write a random beach sunset from the dropdowns alone. The character name overrides (character_1_name, character_2_name) and per-character prompt overrides give you the escape hatch.

What you'd reach for it for

For a storyboard pipeline - several shots of the same two characters in one scene - this node is a nice single point of truth. The multi_ref_scene_prompt output uses the standard image-slot language ("the character in image 1… wearing outfits from image 3 and 4"), so it pairs with the MultiRef_StoryPromptComposer's world, and you still load the reference images yourself. Pair it with the pack's StudioResolutionMaster for consistent canvas, and you've got a coherent starter workflow in three nodes.

Install

The whole pack is pure Python with zero dependencies:

cd ComfyUI/custom_nodes
git clone https://github.com/nexusfinancial-dev/ComfyUI-StudioPromptDirector.git

Or search "ComfyUI-StudioPromptDirector" in ComfyUI Manager and restart. It appears under StudioPromptDirector.

Gotchas

The outputs are only as flexible as the overrides - if the canned video scene fights you, write the Custom_Director_Notes box instead of fighting the dropdowns. Wire char1_casting_prompt / char2_casting_prompt into your Klein text encoder (or into a character-LoRA workflow if you have one), multi_ref_scene_prompt into the multi-ref prompt input, and ltx_video_prompt into the LTX text encoder. And remember the hub is a composer, not a renderer: no model loads here, no images appear - it just writes the words that make the rest of your graph work.