Shorts ads: Google says sound lifts conversions 20%+, add a bed

Google's asset guide for YouTube Shorts ads (read 2026-10-08) tells advertisers to use music, voiceover or both, and says sound has been shown to increase conversions by over 20%. That is Google's statement about its own ads, not a promise for yours. If your Shorts ad is silent, the cheapest test is adding sound in Timeline: a voiceover as the audio spine and a looped music bed as soundtrack with duck_db, which lowers the bed while the voice speaks.

What the page says

The same page asks for vertical 9:16 video, allows up to 3 minutes while only the first 60 seconds play in the Shorts feed, and recommends under 60 seconds. The call-to-action button appears at 3 seconds for Performance Max and Demand Gen campaigns and at 10 seconds for Video view and Video reach campaigns.

Timeline soundtrack fields

audio.url is the spine, for example a voiceover, with duration_seconds as the output length. soundtrack is an optional bed with url, gain_db, loop, fade_out_seconds (up to 10) and duck_db (0 to 20). Ducking needs a real spine: using it with audio.mode: "silence" returns duck_requires_audio_spine. gain_db on the bed is a separate setting, so lower the bed there and use duck for the extra dip under speech.

{
  "audio": {
    "url": "https://media.sume.com/artifacts/artf_demo/voice.wav",
    "duration_seconds": 20
  },
  "video": [
    {
      "source_url": "https://media.sume.com/artifacts/artf_demo/vertical.mp4",
      "start": 0,
      "duration": 20
    }
  ],
  "soundtrack": {
    "url": "https://media.sume.com/artifacts/artf_demo/bed.mp3",
    "gain_db": -12,
    "loop": true,
    "duck_db": 10,
    "fade_out_seconds": 2
  }
}

Testing it fairly

Run the silent and the scored version as two ads with the same targeting and budget, and compare. A 20-second Timeline render is $0.10 at the public rate, so the variant costs ten cents plus the music. Google's figure came from its own data; your lift can be smaller, and for some audiences it is zero. Keep a version with captions, because many viewers still watch Shorts on mute.

Mind the first ten seconds. If the CTA appears at 3 seconds for your campaign type, do not leave the voiceover's key line for later than that.

Level and length notes

A bed at gain_db -12 with duck_db 10 means the music sits 12 dB below the source level and dips another 10 dB while the voice speaks. These numbers are a starting point, not a standard. Listen to a render before you launch, on a phone speaker, since that is where Shorts play.

fade_out_seconds ends the bed smoothly, up to 10 seconds. With loop: true the bed repeats if it is shorter than the spine. Timeline cannot use the bed without a spine when ducking is on, so a music-only ad needs a plain soundtrack with no duck_db.