Edit part of a song with an AI prompt: what to do on Sume

Sume's music docs describe no way to edit part of an existing song. You generate a track from a prompt and get a finished file, so changing one section means writing a new prompt and generating again. Suno announced v6 on its blog on 2026-09-09; this post does not describe what Suno v6 can edit, only what to do on Sume when you want a different bridge or ending.

Can I regenerate only one section?

No route in the music docs takes an existing audio file plus a region. The request fields are prompt, optional image_url, metadata, mode and webhook settings. There is also no seed, temperature, guidance or duration parameter, and duration and duration_seconds are rejected, so a rerun of the same prompt is not a repeat of the same song.

How do I steer a section through the prompt?

The docs suggest section markers such as [0:00-0:30] Intro: ... and a brief with one named moment, for example a breakdown at one time and a full return at another. To move only the ending, keep every other line of the brief word for word and change just the ending clause. Then compare takes and keep the one you like.

Read result.lyrics afterward: the docs call it the model-reported lyrics or section map when present, which is metadata, not an audio measurement.

What does each retry cost?

A fixed $0.125 per accepted Music 1.0 generation, and router models charge the fixed Music price per audio generation. Budget a handful of takes per song rather than one, and store the media URL of any take you want to keep. When you do need separate parts of a mix, see stems.

Which parts of the brief should stay fixed between takes?

Treat the brief as seven axes: emotion, genre, tempo as a number, key and mode, instruments with texture, an arc with one named moment, and era or production. If you like the verse but not the ending, hold six axes exactly as written and rewrite only the arc line. Changing several at once makes it impossible to tell what moved the result.

These are creative directions, not guaranteed output settings, so listen each time.

What is the short version?

Sume cannot edit a section of an existing song. Regenerate with the same brief and a changed arc line, use section markers for structure, and keep the takes you like by their media URLs. Budget several takes per song, since the same prompt does not reproduce the same audio.