Nodes/ComfyUI-Variationator/Batch Variation Generator
ComfyUI Node

Batch Variation Generator

Ten lighting takes on one prompt, in a single run

By MushroomFleet·Created 9 months ago·Updated 5 months ago· 3
Batch Variation Generator
    • var_1
    • var_2
    • var_3
    • var_4
    • var_5
    • var_6
    • var_7
    • var_8
    • var_9
    • var_10
    textSubject description here
    category
    custom_ids1,7,18

    If you've ever wanted to see the same subject at every time of day without hand-editing a prompt ten times, this is the node. Batch Variation Generator takes one base prompt and fans it out into up to ten lighting variations - sunrise through wet night - as list outputs you can run in parallel. It's the pack's answer to A/B testing prompts, and it's the fastest way to learn what a modifier actually does to your images.

    Mechanically it's straightforward. You pick a category and the node resolves that into a set of modifier IDs, then calls the orchestrator's batch generator to append each lighting description to your base prompt. all means IDs 1–20, dry means only the clear-condition modifiers (1–16 for lighting), wet means the low-light/reflection ones (17–20), and custom reads your custom_ids field - a comma-separated string like "1,7,11,18" - so you can build your own day/night cycle: sunrise, golden hour, night, wet blue hour.

    The inputs that matter: text (base prompt), category (all/dry/wet/custom), and custom_ids (only used in custom mode). Outputs are ten STRING list outputs named var_1 through var_10, each a single-item list, so you can feed them to ten separate KSampler branches in parallel.

    Here's the gotcha you need to know before you build that parallel graph: the node always emits all ten outputs, and if your selection produces fewer than ten variations it pads the rest with the original unmodified prompt. So wet gives you 4 real variations and 6 copies of your base text. If you only wire up the first four outputs you're fine, but hooking up all ten silently gives you duplicate renders of the base prompt. That's not a bug in the code - it's the fixed-output-count compromise - but it's easy to burn GPU time on unless you expect it.

    Installing is the pack's standard zero-dependency story: ComfyUI Manager → search "Variationator" → install → restart, or cd ComfyUI/custom_nodes && git clone https://github.com/MushroomFleet/ComfyUI-Variationator.git. No models to download, no requirements.txt to install despite an install-portable.bat that references one. Look under Prompt Orchestration.

    Where this node earns its keep: time-of-day studies for architecture or product shots, and batch tests of a new checkpoint to see how it handles lighting vocabulary. Where it's a trap: it's lighting-only. Despite the generic name, there's no category for mood or weather here - the batch generator is wired to the lighting dimension specifically. And remember the pack's own advice: mixing DRY and WET categories gives you contrast in a comparison set, which is exactly what you want when the goal is "do these variations actually look different?" If they don't, the fix is usually the checkpoint's lighting vocabulary, not the node.

    CategoryPrompt Orchestration

    Inputs (3)

    NameTypeDefaultDescription
    textSTRINGSubject description here
    categoryCOMBO4 options: all, dry, wet, custom
    custom_idsoptSTRING1,7,18

    Outputs (10)

    NameTypeDescription
    var_1STRING
    var_2STRING
    var_3STRING
    var_4STRING
    var_5STRING
    var_6STRING
    var_7STRING
    var_8STRING
    var_9STRING
    var_10STRING