Riffusion alternatives

+ Weights and training code are public - run it locally with no quota.
− Research code: Python and a GPU to set up, and short clips out.

+ Produces radio-ready full songs with sung vocals in any genre within a minute.
− Commercial rights require a paid plan, and it is the subject of major-label copyright litigation.

+ Good at sound effects and samples as well as music, and trained on licensed audio for lower legal risk.
− Tracks top out at a few minutes and vocals are weak compared with Suno or Udio.

+ Licensing is built in - tracks on paid plans are cleared for commercial use.
− Prompt control is coarse next to Suno or Udio; this is background music, not songs.

+ Scores to a timeline and a mood instead of generating a standalone song.
− No vocals, and the range of styles is narrow.

+ Exports MIDI stems so you can edit the composition in a DAW, which prompt-only generators cannot do.
− No vocals and the output can sound generic without manual editing.

+ Clear perpetual license for videos and podcasts, and you can restructure a track's energy section by section.
− Instrumental loops only; it cannot generate vocals or follow a detailed text prompt.

+ Often cleaner, more musical output than Suno with extend and inpaint tools to rework sections.
− Late-2025 label settlements pulled download options, keeping generated songs walled inside the platform.