Nodes/NeuralBooru/NeuralBooru Validator
ComfyUI Node

NeuralBooru Validator

A ComfyUI node in NeuralBooru with 11 inputs and 3 outputs.

By ChrisJohnson89·Created 2 months ago·Updated about a month ago· 15
NeuralBooru Validator
    • prompt
    • dropped_tags
    • tags
    tags
    prompt_templatemasterpiece, best quality, amazing quality, 4k, very aesthetic, high resolution, ultra-detailed, absurdres, newest, scenery, {prompt}, BREAK, depth of field, volumetric lighting
    is_fallbackfalse
    validate_tagstrue
    strict_tagstrue
    fuzzy_cutoff0.00
    min_post_count0
    max_tags0
    exclude_categories
    template_presetcustom
    sort_tagstrue
    CategoryNeuralBooru

    Inputs (11)

    NameTypeDefaultDescription
    tagsSTRINGComma-separated tags to validate. From NeuralBooruLLM or typed manually.
    prompt_templateSTRINGmasterpiece, best quality, amazing quality, 4k, very aesthetic, high resolution, ultra-detailed, absurdres, newest, scenery, {prompt}, BREAK, depth of field, volumetric lightingFinal prompt wrapper; {prompt} is replaced with the validated tags. If {prompt} is missing, tags are appended.
    is_fallbackoptBOOLEANfalseWhen True (set by NeuralBooruLLM on error), skip validation and pass tags through into the template. Reproduces the original node's use_input_text fallback.
    validate_tagsoptBOOLEANtrueCheck every tag against the real Danbooru vocabulary.
    strict_tagsoptBOOLEANtrueDrop candidates that match no real tag. Off keeps them raw.
    fuzzy_cutoffoptFLOAT0.000–1Remap near-miss tags by similarity (0.8+ recommended). 0 disables fuzzy matching.
    min_post_countoptINT00–1000000Drop tags with fewer Danbooru posts than this. 0 keeps all.
    max_tagsoptINT00–200Keep at most this many tags. 0 means no limit.
    exclude_categoriesoptSTRINGTag categories to drop, comma-separated: artist, character, copyright, meta, general.
    template_presetoptCOMBOcustomQuality-tag wrapper for common model families. custom uses the prompt_template field above.
    sort_tagsoptBOOLEANtrueReorder tags Danbooru-style: people counts, character, copyright, artist, general, meta. Booru-trained models respond to this order.

    Outputs (3)

    NameTypeDescription
    promptSTRINGValidated tags wrapped in the prompt template, for your CLIP encoder.
    dropped_tagsSTRINGCandidates that failed validation, comma-separated.
    tagsSTRINGThe validated tags alone, without the template.