Doom Constructor

Doom Constructor is for the days you stare at an empty prompt box and draw a blank. It replaces that box with nine dropdowns - one per prompt category - and assembles your prompt from whatever you pick. Colors, materials, styles, artists, lighting, camera angle, mood, quality keywords, era. Pick a few, skip the rest, get a comma-joined prompt out the other side. It's a structured alternative to typing, and for seed-farming prompts it's surprisingly effective at forcing variety you'd never type by hand.

It comes from the pack's prompt-engine family, the same lineage as Doom Action Prompter and Doom Edit Prompts (this one's old name was KontextConstructor_RU). The output is a plain string, so it plugs into any CLIP text encode or directly into a sampler's prompt field.

How it works

Nine combo inputs, one per category: Цвета (colors), Материалы и Текстуры, Стили и Жанры, Художники и Студии, Освещение и Время суток, Ракурс и Композиция, Эмоции и Настроение, Детализация и Качество, Эпохи и Культуры. Each defaults to "не выбрано" (not selected), and any category left on that default is skipped entirely - no stray "не выбрано" text leaks into your prompt.

Pick, say, "golden hour" from lighting, "dutch angle" from composition, "epic" from mood, and "highly detailed, 8K" from quality, and the single output, prompt_parts, gives you:

golden hour, dutch angle, epic, highly detailed, 8K

That's the whole mechanism. It joins your chosen values with commas and hands them over. No model, no API, nothing to break.

The built-in vocabulary is English and largely tuned for the current generation of models - "Ghibli Studio", "cyberpunk", "cinematic lighting", "4K/8K" quality tags are the kind of thing the 2026 LLM-encoder models actually respond to. Artists lean western and a bit dated (Van Gogh, Dali, Mucha, Greg Rutkowski, Artgerm) but they still work as style anchors.

Why you'd reach for it

Two reasons. First, it makes prompts reviewable: you can see at a glance what's in them, unlike a wall of free text. Second, it's a great companion to a seed loop - swap one dropdown value per run and you get clean, comparable variations without retyping a sentence each time.

Installing it

Same as every node in this pack - there's no separate install:

cd ComfyUI/custom_nodes
git clone https://github.com/PeterMikhai/Doom_Flux_NodePack

Restart ComfyUI, or grab the pack through ComfyUI Manager. No extra dependencies. The README's old clone URL (DoomAI_nodes.git) points at a renamed repo; the current one is Doom_Flux_NodePack.

Common issues

• Prompt looks short or generic. The defaults are all off, so if you only flip one or two dropdowns you get a one-or-two-tag prompt. That's the node working as designed - but it means a half-configured Constructor produces a weaker prompt than the average typed one. Load it up or don't bother.
• Russian labels. All nine category headings are Russian ("Ракурс и Композиция" is composition, "Эпохи и Культуры" is eras). The values themselves are English. If you don't read Russian, keep a mental map of the last three: emotions, detail/quality, eras.
• Legacy workflows. Old graphs referencing KontextConstructor_RU still load - the alias is registered for backward compatibility.

It's not the node that changes your life, but it's the node that changes your prompts from unexamined habit into a set of knobs you can actually tweak.