WCAG 1.4.7: Low or No Background Audio
A music bed that sounds “subtle” to a hearing producer can bury every consonant for a hard-of-hearing listener. For prerecorded speech audio, this AAA criterion requires no background sound, a way to turn it off, or a background at least 20 dB below the speech — roughly four times quieter.
The success criterion, in full
For prerecorded audio-only content that (1) contains primarily speech in the foreground, (2) is not an audio CAPTCHA or audio logo, and (3) is not vocalization intended to be primarily musical expression such as singing or rapping, at least one of the following is true: No Background — the audio does not contain background sounds; Turn Off — the background sounds can be turned off; 20 dB — the background sounds are at least 20 decibels lower than the foreground speech content, with the exception of occasional sounds that last for only one or two seconds.
Per the accompanying note, a 20 dB difference means the background is approximately four times quieter than the foreground speech.
What is in scope — and what is not
In scope
- Podcasts and narrated audio articles
- Recorded lectures, briefings, and interviews published as audio
- Audio tours and guides
- Any prerecorded, audio-only file whose point is the speech
Out of scope
- Video soundtracks (synchronized media)
- Live audio streams
- Audio CAPTCHAs and audio logos
- Singing and rapping — vocals as musical expression
This is the AAA companion to 1.4.2 Audio Control (Level A), which handles auto-playing audio, and to 1.2.1, which requires a transcript for the same content. The transcript is the fallback; 1.4.7 is about making the audio itself usable for people who want to listen.
Who this helps
People who are hard of hearing
Hearing loss sharply reduces the ability to separate speech from competing sound. High-frequency loss in particular erases consonants first — exactly what background music masks.
Hearing aid and CI users
Hearing aids and cochlear implants compress the dynamic range, which pushes background sound closer to speech. A generous margin in the source mix survives that processing.
People with auditory processing differences
Some listeners hear at normal thresholds but cannot filter overlapping streams of sound. Clean speech dramatically reduces the effort of listening.
Non-native listeners and noisy environments
Understanding a second language, or listening on transit through cheap earbuds, both consume the same headroom the background music is eating.
Pass and fail examples
✓ Passes
- A narrated article recorded as clean voice with no bed at all.
- A podcast whose intro sting plays, ends, and then leaves the conversation dry.
- A documentary-style piece mixed with the ambience bed 24 dB under the narration (RMS).
- A player offering “voice only” and “full mix” versions of the same episode.
✗ Fails
- A guided meditation with continuous music only ~8 dB under the voice.
- An interview recorded in a café, espresso machine competing with the answers.
- A produced piece where music swells over the narrator at every section break for 10+ seconds.
- “Turn off music” exists — but only as a setting that re-renders future episodes, not this one.
Implementation patterns
1. Offer a voice-only alternative (“Turn Off” branch)
Export two masters from the same session — full mix and narration only — and let the listener choose.
<h2>Episode 42: Designing for everyone</h2>
<audio controls src="/audio/ep42-full-mix.mp3"></audio>
<p>
<a href="/audio/ep42-voice-only.mp3">
Listen to the voice-only version (no background music)
</a>
· <a href="/transcripts/ep42">Read the transcript</a>
</p>2. A real-time background toggle with the Web Audio API
If you ship speech and background as separate tracks, the “off” switch can live in the player itself.
<audio id="speech" src="/audio/ep42-voice.mp3" controls></audio>
<audio id="bed" src="/audio/ep42-music.mp3" muted></audio>
<label>
<input type="checkbox" id="bed-toggle" />
Play background music (at reduced volume)
</label>
<script>
const speech = document.getElementById("speech");
const bed = document.getElementById("bed");
const toggle = document.getElementById("bed-toggle");
bed.volume = 0.1; // ≈ -20 dB relative to full scale
toggle.addEventListener("change", () => {
bed.muted = !toggle.checked;
});
// Keep the two elements locked together
speech.addEventListener("play", () => { bed.currentTime = speech.currentTime; bed.play(); });
speech.addEventListener("pause", () => bed.pause());
speech.addEventListener("seeked", () => { bed.currentTime = speech.currentTime; });
</script>3. Production checklist for the 20 dB branch
- Meter speech and background separately using RMS or LUFS, not peak: if narration averages -16 LUFS, keep the bed at or below -36 LUFS.
- Duck the bed during speech rather than relying on a fixed level; automation should hold the 20 dB margin whenever anyone talks.
- Reserve louder moments for gaps in speech or keep them under about two seconds — the criterion's 'occasional sounds' allowance.
- Check the final encode on laptop speakers and earbuds; heavy MP3 compression and small drivers can smear a marginal mix.
How to test for 1.4.7
- 1
Identify in-scope audio
List prerecorded, audio-only content where speech is the primary content. Set aside video soundtracks, live streams, CAPTCHAs, audio logos, and musical vocals.
- 2
Listen for background sound
Play each file with decent headphones. If there is genuinely no background sound, the file passes on the 'No Background' branch and you are done.
- 3
Look for a turn-off mechanism
If background exists, check for a voice-only version, a stem toggle in the player, or an equivalent published alternative — and confirm it actually removes the background.
- 4
Measure the level difference
Otherwise, load the audio into an editor. Compare speech-heavy passages against background-only passages (or stems) using RMS/LUFS. The background must sit at least 20 dB below the speech.
- 5
Apply the occasional-sounds allowance correctly
Brief sounds of one or two seconds may exceed the limit. Sustained beds, long swells, and recurring loud effects may not — flag anything that competes with speech for longer than a moment.
Common failures
- A continuous music bed mixed at 'production-tasteful' levels — typically 8–12 dB under speech, half the required margin.
- Recording interviews in noisy rooms, baking irreducible background into the only master that exists.
- Music swells over narration at transitions that run far past the one-to-two-second allowance.
- Judging by ear on studio monitors only — a mix that seems fine to a hearing engineer at high volume fails listeners with hearing loss on earbuds.
- Offering a 'voice only' link that points to the same full-mix file (broken alternates are surprisingly common).
- Assuming the transcript excuses the audio — 1.2.1's transcript is a separate requirement, not a substitute for a usable mix.
Frequently asked questions
What does WCAG 1.4.7 Low or No Background Audio require?
For prerecorded audio-only content that primarily contains speech in the foreground, at least one of the following must be true: the audio contains no background sounds; the background sounds can be turned off; or the background sounds are at least 20 decibels lower than the foreground speech, except for occasional sounds lasting only one or two seconds. Content that is an audio CAPTCHA, an audio logo, or vocalization intended primarily as musical expression (singing, rapping) is out of scope. It is a Level AAA criterion from WCAG 2.0.
How loud is a 20 dB difference in practice?
Decibels are logarithmic, so 20 dB is a large gap: the background is about one-quarter as loud as the speech perceptually (and one-tenth the sound pressure). The W3C's own note puts it plainly — background sound that meets this requirement will be approximately four times quieter than the foreground speech. If you can comfortably follow every word without straining to separate it from the music bed, you are probably in the right territory; verify with metering rather than ears alone.
What content does 1.4.7 apply to — and what is excluded?
It applies to prerecorded audio-only content whose primary content is speech: podcasts, narrated audio articles, recorded lectures and briefings, audio guides. It does not apply to video soundtracks (audio that is part of synchronized media is out of this criterion's scope), to live audio, to audio CAPTCHAs and audio logos, or to performances where the vocals are the music — a song is allowed to have instruments behind the singer.
Why do background sounds matter so much for hard-of-hearing listeners?
People who are hard of hearing have a substantially reduced ability to separate speech from competing sound — a skill hearing people exercise unconsciously. Around 1 in 12 people have some hearing loss, and for many of them a music bed at 'tasteful' production levels turns speech into mush, especially with high-frequency loss where consonants live. Cognitive processing differences and auditory processing disorder create similar difficulty. A 20 dB margin restores the separation that typical mixing assumes.
How do I measure whether my mix meets the 20 dB rule?
In your audio editor (Audacity, Adobe Audition, Reaper), compare the level of a speech-only section against a background-only section — or better, export the stems. Measure with RMS or LUFS rather than instantaneous peaks, since perceived loudness is what matters. If speech averages, say, -16 LUFS, the background bed should sit at or below -36 LUFS. Occasional one-to-two-second sounds (a transition sting, a door closing) may exceed the limit without failing.
What is the easiest way to conform?
Record and publish clean narration — no background sounds means automatic conformance, and it is what most listeners prefer for spoken content anyway. If the production style calls for music, publish an alternate 'voice-only' version or deliver background as a separately toggleable track, which satisfies the 'can be turned off' branch. Mixing to the 20 dB spec is the third route, and the one requiring actual metering discipline.
Related Success Criteria
Color is not used as the only visual means of conveying information.
Audio that plays automatically can be paused, stopped, or controlled.
Text has a contrast ratio of at least 4.5:1 (3:1 for large text).
Text can be resized up to 200% without loss of content or functionality.
Text is used instead of images of text, except for customizable or essential images.