Doom Action Prompter

Doom Action Prompter is the pack's "I know what I want to do to this image, I just can't be bothered phrasing it" node. It's a dropdown of 20 image-editing commands - stylize, swap the background, change the lighting, replace an object, fix the composition, rewrite some text - and it hands you back a ready-to-use English prompt. No API, no model, no key. It's pure string substitution, which is exactly why it's dependable: it can't hallucinate a template.

It's part of the prompt-engine family in this pack (Constructor, Edit Presets, Wordbook, all of it formerly the Russian "Kontext" line - this node's old name was KontextPrompterUltimate_RU). It targets the img2img / edit workflows you'd build with the pack's DoomFlux2I2ISampler or DoomFlux1InpaintSampler, but the output is a plain string, so it feeds any CLIP text encode or edit sampler anywhere.

How it works

A combo input, action_choice, is filled with entries formatted as Category >> action - everything from "Реставрация и Улучшение >> Раскрасить Ч/Б фото" to "Текст >> Заменить текст". Here's the mildly surprising part: the dropdown labels are Russian, but the actual prompts they emit are English. Pick "Стилизация >> Применить стиль" and you get turn this photo into {input1} style.

Two string inputs, input_1 and input_2, fill the {input1} / {input2} slots in whichever template you chose. Not every template has both slots - "Add a {input1} to the scene" only uses input 1, "Change the color of the {input1} to {input2}" uses both. Leave unused ones empty; they're just ignored. The single output, prompt, is the assembled command, e.g. Change the color of the car to red.

Why you'd reach for it

Prompt consistency. When you're batch-editing, a human retypes "make it Ghibli-style" six slightly different ways across the day; this node can't. Pick the template, type the one variable thing, and every render gets identical phrasing. That's genuinely useful for the seed-farming workflow where you want to change one variable at a time and nothing else.

Installing it

The whole pack installs as one unit - there's no standalone download for this node:

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

Restart ComfyUI. No extra pip dependencies; the pack runs on ComfyUI's own stack (torch, PIL, aiohttp). You can also use ComfyUI Manager and search for the pack. One wrinkle: the README still shows the old repo name DoomAI_nodes.git - it was renamed, and the current clone URL is Doom_Flux_NodePack.

Common issues

• Russian dropdown, English output. If you pick a template and the prompt looks nothing like the label, that's expected. The label is a description; the emitted prompt is the actual edit instruction.
• Forgetting the slot. "Replace the text {input1} with {input2}" with both inputs empty produces a prompt with literal {input1} braces in it. That's a prompt you probably don't want. If a template has slots, fill them.
• Old workflow referencing KontextPrompterUltimate_RU? Fine. The old class name is still registered as an alias, so legacy graphs load without remapping.

It's a small node with a narrow job, and it does that job well. If you don't edit images in batches, you won't miss it. If you do, it's the kind of thing that quietly saves you from retyping the same command for the fortieth time.