Nodes/ComfyUI-SongScribe/Style Preset (SongScribe)
ComfyUI Node

Style Preset (SongScribe)

Generate a style prompt from a curated preset, for MiniMax Music 3 or YuE2. Presets are YAML files in songscribe/presets - drop in your own and they appear here on the next restart.

By TheLocalLab·Created about a month ago·Updated a day ago· 11
Style Preset (SongScribe)
    • prompt
    • global_metadata
    • vocal_details
    • arrangement
    presetAcoustic / Bluegrass
    formatminimax
    detailfull
    vocalauto
    languageauto
    stylebalanced
    eranone
    texturenone
    mood_shiftnone
    seed0
    blend_withnone
    blend0.50
    extra
    CategorySongScribe

    Inputs (13)

    NameTypeDefaultDescription
    presetCOMBOAcoustic / Bluegrass73 options: Acoustic / Bluegrass, Acoustic / Country, Acoustic / Indie Folk, Acoustic / Solo Piano, Cinematic / Classical Chamber, Cinematic / Epic Trailer, +67
    formatCOMBOminimaxminimax: three labelled sections (Global Metadata / Vocal Details / Arrangement). yue2: one flat comma-separated descriptor string, which is what YuE2's style input expects.
    detailCOMBOfullYuE2 only. tags: genre, tempo, mood and vocal. full: adds instrumentation and production. rich: adds scene and extra texture. YuE2's own prompts range across all three.
    vocalCOMBOautoYuE2 states vocal type explicitly in most prompts. 'auto' uses whatever the preset declares.
    languageCOMBOautoSung language. YuE2 supports English, Mandarin, Japanese, Korean, Spanish and Russian; naming it steers pronunciation.
    styleCOMBObalancedMiniMax only: how much measured detail the caption carries.
    eraCOMBOnoneLayer a production era on top.
    textureCOMBOnoneLayer a recording texture on top.
    mood_shiftCOMBOnonePush the mood in a direction.
    seedINT00–18446744073709550000
    blend_withoptCOMBOnoneOptional second preset to mix in.
    blendoptFLOAT0.500–1How much of the second preset to admit. 0 = first only, 1 = second only.
    extraoptSTRINGAppended verbatim. Use for anything the preset cannot know - a reference artist, a negative instruction like 'avoid heavy distortion', or a specific structure.

    Outputs (4)

    NameTypeDescription
    promptSTRINGThe style prompt, in the selected format. Wire to YuE2's style input or MiniMax's caption input.
    global_metadataSTRINGMiniMax section 1 (empty detail in yue2 format is still filled).
    vocal_detailsSTRINGMiniMax section 2.
    arrangementSTRINGMiniMax section 3.