Are FUZZ-1.0 Pro and FUZZ-1.0 the same model?
No. They are separately listed models, and FUZZ-1.0 Pro belongs to the historical Pro tier. When calling it, explicitly specify FUZZ-1.0 Pro to avoid relying on the default selection. When comparing the two, it is recommended to listen using the same lyrics and musical direction rather than judging the final result solely by the Pro name.
Is a single musical idea enough, or must I prepare complete lyrics?
Creation can start from a text concept, or you can use custom lyrics. In the request structure, /riffusion/audios requires action and lyric, while /producer/audios also requires prompt; inspiration mode must retain required fields as well, so you cannot directly copy simplified requests that omit fields.
How do I continue a song generated by FUZZ-1.0 Pro?
Save the song result's id, use it as audio_id in the continuation request, set action to extend, and use continue_at to specify the position from which to continue, in seconds. Also prepare continuation lyrics or a musical direction, then listen to the transition between the old and new content after completion.
Can I modify only one section of lyrics in a song?
You can use the replace_section workflow, submitting the song's audio_id, the replacement lyrics, and the time range in seconds specified by replace_section_start and replace_section_end. This regenerates the specified segment rather than replacing an existing recording word by word, so check the connections before and after the change.
Can a generation task be completed in the background and retrieved later?
Yes. After setting async=true, first obtain task_id, then query the task status; you can also configure callback_url to receive the completed result. The application should save the task ID, read audio_url in data once the result is ready, and save the song ID for later covers, continuations, or partial modifications.
