WCAG 1.4.7 — Low or No Background Audio
Spoken audio with music underneath sounds professional. For users with auditory processing difficulties or moderate hearing loss, that same background track turns the speech into noise.
What this requires
For prerecorded audio-only content that is primarily speech, the audio must satisfy one of three conditions: no background sounds, the background sounds can be turned off, or the background sounds are at least 20 dB lower than the foreground speech (roughly four times quieter, perceptually). This is the AAA "is the podcast actually intelligible to people with mild hearing loss?" criterion.
How AI coding tools fail this
Code-generation tools don't mix audio, so the criterion's main failure sits in production. Where AI does intersect: AI-generated voiceovers combined with AI-generated background music tracks tend to be mixed hot on both, with no quiet pass. The result is a clean-sounding demo that fails the 20 dB requirement.
Code-side, the related failure is omitting a mute control for any "background music" on speech-bearing pages. Even when the audio file itself is fine, layering a background ambient loop on a podcast page puts both sounds at the same level and degrades the speech.
Edge cases
- The 20 dB requirement is roughly "four times quieter". A decent audio engineer can mix to this comfortably; an AI music generator without a quiet pass usually can't.
- Captions and transcripts don't satisfy this criterion. 1.4.7 is about the audio being intelligible by ear, not about an alternative for non-hearing users.
- Brief musical introductions that fade out before the speech begins are fine — the criterion targets simultaneous background audio.
- Music podcasts where the music is the point are not in scope (the speech is incidental to the music).
- AAA is rare for audio-heavy properties because production cost scales with it. Plan it into the editorial workflow if you target it.
How Jeikin handles this
This is a manual review of audio mastering — not something static
analysis can catch. The dashboard lists every <audio> element on
the site and asks the reviewer to confirm the mix meets the
threshold or that an alternative speech-only version is offered. The
evidence is the reviewer's confirmation.