Sign in
Recipes

Generates full songs from text descriptions and structured lyrics with ACE Step 1.5.

Changed on
post/v2/consumer/recipes/aceStepAudio

Request

  • The document declares no server URL.
  • Auth: none required

Query parameters

experimentalboolean
allowMatureContentboolean
whatifboolean
ephemeralboolean
tier'paid' | 'free'

How a workflow is paid for.

Request body

musicDescriptionstring required

Music style/genre description (e.g., "Neo-Soul: A warm, organic neo-soul track...")

lyricsstring required

Structured lyrics with section markers like [Verse], [Chorus], [Bridge], etc.

seedinteger required
durationnumber float

Duration in seconds (1-190)

bpminteger

Beats per minute (40-200)

timeSignature'2' | '3' | '4' | '6'

Time signature (e.g., "4" for 4/4 time)

language'en' | 'ja' | 'zh' | 'es' | 'de' | 'fr' | 'pt' | 'ru' | 'it' | 'nl' | 'pl' | 'tr' | 'vi' | 'cs' | 'fa' | 'id' | 'ko' | 'uk' | 'hu' | 'ar' | 'sv' | 'ro' | 'el'

Language code (e.g., "en", "zh", "ja", "ko")

key'C major' | 'C# major' | 'Db major' | 'D major' | 'D# major' | 'Eb major' | 'E major' | 'F major' | 'F# major' | 'Gb major' | 'G major' | 'G# major' | 'Ab major' | 'A major' | 'A# major' | 'Bb major' | 'B major' | 'C minor' | 'C# minor' | 'Db minor' | 'D minor' | 'D# minor' | 'Eb minor' | 'E minor' | 'F minor' | 'F# minor' | 'Gb minor' | 'G minor' | 'G# minor' | 'Ab minor' | 'A minor' | 'A# minor' | 'Bb minor' | 'B minor'

Musical key (e.g., "C major", "E minor"). Mirrors ComfyUI's TextEncodeAceStepAudio1.5 keyscale combo: 17 roots × {major, minor}.

instrumentalWeightnumber float

Weight for instrumental elements (0.0-1.0)

vocalWeightnumber float

Weight for vocal elements (0.0-1.0)

diffusionModelstring nullable

Optional diffusion model (unet) override. Defaults to ACE-Step 1.5 turbo 2B.

clipModelstring nullable

Optional text encoder (CLIP) override. Defaults to qwen_4b_ace15.

vaeModelstring nullable

Optional VAE override. Defaults to ace_1.5_vae.

languageModelstring nullable

Optional language model override for audio code generation. Defaults to qwen_0.6b_ace15.

stepsinteger

Number of sampling steps. Turbo variants converge at 8; non-turbo base / sft variants expect the full 50-step schedule.

cfgnumber float

Classifier-free guidance scale. Turbo variants run CFG-off at 1.0; non-turbo base / sft variants expect CFG on (around 4).

lorasobject

LoRAs to apply; each strength applies to both the diffusion model and the dual CLIP. Compatibility with non-default base models is the caller's responsibility.

Response

OK

Changes