Nodes/crypswolf69/✦ ALCHEMIST // DATASET CAPTIONER
ComfyUI Node

✦ ALCHEMIST // DATASET CAPTIONER

Walk a folder and write a same-stem .txt caption beside each image, using the Krea 2 LoRA training recipe. Set overwrite off to only fill in the gaps.

By babydjac·Created 5 days ago·Updated 5 days ago· 0
✦ ALCHEMIST // DATASET CAPTIONER
    • report
    • captioned
    • skipped
    • failed
    ◄actionidle►
    ◄folder/workspace/Pawg TS►
    ◄providerxAI / Grok►
    ◄modelgrok-4.20-non-reasoning►
    ◄reasoningminimal►
    ◄lora_kindbody concept►
    ◄trigger_word►
    ◄dry_runtrue►
    ◄limit0►
    ◄overwritefalse►
    ◄nsfwtrue►
    ◄caption_words45►
    ◄omit_terms►
    ◄extra_direction►
    ◄api_key►
    ◄temperature0.70►
    ◄custom_base_url►
    Categoryprompting/alchemist

    Inputs (17)

    NameTypeDefaultDescription
    actionCOMBOidleRuns in the browser and resets to idle. Never queues a generation.
    folderSTRING/workspace/Pawg TSFolder of images. Captions land next to them as .txt.
    providerCOMBOxAI / GrokWhich LLM answers. Use FETCH MODELS after switching.
    modelCOMBOgrok-4.20-non-reasoningModel id. FETCH MODELS repopulates this from the live API.
    reasoningCOMBOminimalPaid per image here, so keep it low.
    lora_kindCOMBObody concept4 options: character, body concept, style, clothing or object
    trigger_wordSTRING—
    dry_runBOOLEANtrue—
    limitINT00–50000 = no cap. Set 3 for a cheap sanity check first.
    overwriteBOOLEANfalse—
    nsfwBOOLEANtrue—
    caption_wordsoptINT4520–70—
    omit_termsoptSTRING—
    extra_directionoptSTRING—
    api_keyoptSTRING—
    temperatureoptFLOAT0.700–2—
    custom_base_urloptSTRING—

    Outputs (4)

    NameTypeDescription
    reportSTRING—
    captionedINT—
    skippedINT—
    failedINT—