AI sound effects: Suno Sounds vs a music prompt

Suno Sounds V5.5 is a sound generation model priced at 2 credits flat, according to a Versely roundup. Sume has no dedicated sound-effects model. The closest route is the Music Router with a prompt that asks for no vocals; it returns audio, but the docs promise a music engine, not effects, so listen before you ship.

What Versely reports

From the Versely state of AI voice post, read on 2026-10-03:

What Sume offers

Sume documents music generation through the Music Router, where sume/music-auto resolves to Lyria 3.5 today. Speech is a separate tool (tts_create on MCP). There is no sound-effects model in the docs, so any effect you get comes from steering a music engine with text.

A prompt that asks for ambience

The Music prompt guidance says to close with one clause such as "Instrumental, no vocals." and to put exclusions in the positive prompt, since a non-empty negative_prompt is unsupported. Length is steered in the prompt because duration and duration_seconds are rejected. Provider lyrics metadata is model-reported, not an audio measurement.

curl -X POST https://api.sume.com/v1/music-router/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: sfx-door-001" \
  -d '{
    "model": "sume/music-auto",
    "prompt": "A 10-second ambient texture: a heavy wooden door creaking open in an empty hall, low room tone, no melody. Instrumental, no vocals."
  }'

How the two differ in practice

A sound model is trained to produce events such as hits, whooshes or footsteps. A music engine is trained to produce songs, so the output may add rhythm or harmony you did not ask for. The docs call the music brief axes creative directions, not guaranteed output settings, and tell you to verify the generated audio.

• Generate two or three takes and keep the one that is closest to a plain effect.
• Trim with the video and audio tools you already use, or lay the file under footage in Timeline.
• If you need a specific library sound, use a source built for effects and bring the file in as an audio input.