ace-step

Text-to-audio generation creates instrumental or vocal tracks from genre, mood, instrument, tempo, and lyric tags. You can produce background-music drafts, multilingual vocal songs, or game loops by providing structured JSON inputs to the model.

Audio inpainting regenerates a specific time range in an existing track to fix a chorus, replace a bridge, or revise another section. It requires HTTPS source audio and endpoint selection based on your inpainting intent.

Audio outpainting extends tracks before or after the original recording to add intros, outros, fades, or longer arrangements. This allows you to create a polished longer cut from a short musical hook.

Yes, using the ACE Step models requires the RunComfy CLI and authentication. You must provide structured JSON inputs and select the appropriate endpoint based on whether your intent is generation, inpainting, or outpainting.

Yes, text-to-audio generation supports multilingual lyrics for vocal composition. You can specify lyric tags alongside genre, mood, instrument, and tempo tags to create inexpensive vocal tracks.

Audio inpainting requires HTTPS source audio for editing workflows and is limited to regenerating specific time ranges. You must select the correct endpoint and provide structured JSON inputs to execute the repair successfully.