ACE-Step Cover

A ComfyUI node in Audio/ACE-Step with 22 inputs and 3 outputs.

By kana112233·Created 6 months ago·Updated 6 months ago· 27

ACE-Step Cover

◄caption►

◄checkpoint_dirAce-Step1.5►

◄config_pathacestep-v15-turbo►

◄lm_model_pathacestep-5Hz-lm-1.7B►

◄audio_cover_strength0.50►

◄batch_size1►

◄seed-1►

◄inference_steps8►

◄guidance_scale7.00►

◄deviceauto►

◄lyrics►

◄vocal_languageunknown►

◄instrumentalfalse►

◄bpm0►

◄keyscale►

◄timesignature►

◄use_adgfalse►

◄thinkingtrue►

◄audio_formatflac►

◄instruction►

CategoryAudio/ACE-Step

Inputs (22)

Name	Type	Default	Description
src_audio	AUDIO		Source audio to be covered (remade in new style).
caption	STRING		Description of the target musical style (e.g., 'A jazz version of this song').
checkpoint_dir	COMBO	Ace-Step1.5	Directory containing ACE-Step model weights.
config_path	COMBO	acestep-v15-turbo	Model configuration (turbo is faster).
lm_model_path	COMBO	acestep-5Hz-lm-1.7B	Language model for metadata/lyrics.
audio_cover_strength	FLOAT	0.500–1	Strength of preserving original audio structure. Use LOWER (0.1-0.3) for more style change.
batch_size	INT	11–4	Number of variations.
seed	INT	-1-1–18446744073709550000	Random seed.
inference_steps	INT	81–100	Steps. Turbo: 8, Base: 50+.
guidance_scale	FLOAT	7.001–15	CFG strength.
device	COMBO	auto	Processing platform.
lyricsopt	STRING		Optional lyrics.
vocal_languageopt	COMBO	unknown	Target language.
instrumentalopt	BOOLEAN	false	Instrumental mode.
bpmopt	INT	00–300	Target BPM (0 for keep original).
keyscaleopt	STRING		Musical key.
timesignatureopt	STRING		Time signature.
use_adgopt	BOOLEAN	false	Adaptive Dual Guidance.
thinkingopt	BOOLEAN	true	Show LLM reasoning.
audio_formatopt	COMBO	flac	Output format.
lora_infoopt	ACE_STEP_LORA_INFO		Optional LoRA style model.
instructionopt	STRING		Custom instruction (overrides default cover instruction).