Prompt Nodes

March 21, 2026 ยท View on GitHub

The Prompt category handles text generation, text formatting, and native ACE-Step text encoding. These nodes prepare the semantic and structural guidance before passing it to the sampler or conditioning pipeline.

1. ScromfyAceStepTextEncoderPlusPlus

File: nodes/text_encoder_plusplus_node.py

The definitive ACE-Step 1.5 text encoder. Combines SFT "Enriched CoT" (Chain-of-Thought) formatting with granular base controls. Supports a toggle for enhanced prompting, trigger words/tags, and full LLM parameters.

Inputs

  • clip (Required, CLIP): The loaded ACE15TEModel.
  • caption (Required, STRING): The core description (e.g. genre, mood, instruments).
  • enhanced_prompt (Required, BOOLEAN): If True, uses SFT-style 'Enriched CoT' (YAML) formatting for better fine-tuned model performance. If False, uses native ComfyUI encoding logic.
  • generate_audio_codes (Required, BOOLEAN): Enable LLM audio code generation (semantic structure).
  • instrumental (Required, BOOLEAN): Force [Instrumental] as lyrics.

Optional Advanced Overrides: lyrics, trigger_word, style_tags, negative_prompt, bpm (-1 = auto), duration, keyscale, timesignature, language. Optional LM Sampling: seed, cfg_scale, temperature, top_p, top_k, min_p, repetition_penalty.

Outputs

  • conditioning (CONDITIONING): The encoded positive conditioning data.
  • zero_conditioning (CONDITIONING): Automatically zeroed-out conditioning suitable for negative input.

2. AceStepMetadataBuilder

File: nodes/metadata_builder_node.py

Format raw integers, floats, and strings into a strict music metadata dictionary for Ace-Step conditioning manipulation.

Inputs

  • bpm (INT): 0-300.
  • duration (FLOAT): Length in seconds.
  • keyscale (STRING): E.g. "C major".
  • timesignature (INT): E.g. 4.
  • language (STRING): Country code drop-down.
  • instrumental (BOOLEAN).

Outputs

  • metadata (DICT): A clean, filtered dictionary mapping these key-values (omitting empty defaults) to pass into other conditioning nodes.

3. ScromfyAceStepPromptGen

File: nodes/prompt_gen_node.py

Dynamic multi-category prompt generator using weighted tags. It scans the prompt_components/ directory to expose dropdown lists of musical features (genres, moods, instruments, setups) and allows for specific selection or randomization.

Inputs

  • Dynamic Text Lists (Required): All .txt files in prompt_components appear as dropdowns. Options include specific elements, none, random, or random2.
  • seed (Required, INT): Deterministic seed for random picks.

Outputs

  • combined_prompt (STRING): The full concatenated master string.
  • [category]_text (STRING): A separate output port is automatically generated for every component category evaluated (e.g. genres_text).

4. AceStepRandomPrompt

File: nodes/random_prompt_node.py

Generate random music prompts utilizing the predefined sentence templates provided in prompt_components/lyrics/SONG_PROMPT_TEMPLATES.txt.

Inputs

  • seed (Required, INT).
  • template (Required, DROP-DOWN): List of specific structural templates or completely random.

Outputs

  • prompt (STRING): The finalized randomly generated sentence string.

5. Prompt Freeform

File: nodes/prompt_freeform_node.py

Resolves __WILDCARDS__ in any chunk of freeform text into dynamic components using files found in prompt_components/.

Inputs

  • text (Required, STRING): E.g., "A fast GENRE song with heavy INSTRUMENTS".
  • seed (Required, INT): Deterministic resolution.

Outputs

  • text (STRING): The fully resolved prompt, substituting the wildcards with parsed data.