arioso.platforms.ace_step.adapter#

ACE-Step adapter (fal.ai): text-to-music and audio-to-audio remix.

ACE-Step is steered by comma-separated style tags, not a sentence, so the unified prompt (or genre) becomes tags. Its audio-to-audio endpoint “remixes” an input toward new tags; it also needs original_tags describing the input, which default to the new tags when not given.

Classes

Adapter(config)

fal.ai ACE-Step adapter.
class arioso.platforms.ace_step.adapter.Adapter(config)[source]#

Bases: object

fal.ai ACE-Step adapter.

generate(prompt='', *, genre='', lyrics='', instrumental=False, duration=60.0, num_steps=27, guidance=15.0, seed=None, audio_input=None, original_tags='', original_lyrics='', edit_mode='remix', fetch=True, **kwargs)[source]#

Generate music from tags, or remix audio_input toward them.

Parameters:

  • prompt (str) – Style tags (comma-separated works best). genre wins when both are given.

  • genre (str) – Style tags; an alias of prompt for this platform.

  • lyrics (str) – Lyrics to sing. Empty or instrumental=True sends [inst], so a remix of a song with vocals comes out instrumental unless you pass its lyrics (or new ones).

  • instrumental (bool) – Force an instrumental (discards lyrics).

  • duration (float) – Seconds (text-to-music only; a remix keeps the input’s length).

  • num_steps (int) – number_of_steps (default 27).

  • guidance (float) – guidance_scale (default 15).

  • seed (int) – Random seed.

  • audio_input – Audio to remix (path, bytes, Song, URL, …). When given, the audio-to-audio endpoint is used.

  • original_tags (str) – Tags describing audio_input (defaults to the new tags).

  • original_lyrics (str) – Lyrics of audio_input, if any.

  • edit_mode (str) – "remix" (default) or "lyrics".

  • fetch (bool) – Download the result so audio_bytes is populated.

  • **kwargs – Native fal arguments in _NATIVE_EXTRAS (scheduler, guidance_type, tag_guidance_scale, …) are passed through; anything else is ignored.

Return type:

Song

Returns:

A completed Song.