The <audio> Element

Technical Summary

Definition / Categories

Item	Definition
Meaning	Plays a media resource whose data is ostensibly audio in a document
Categories	Flow, phrasing, and embedded content. With controls, it is also interactive and palpable content
Context	Where embedded content is expected
Content model	The presence of src changes whether child source elements are allowed and their order. Both forms allow zero or more track elements followed by transparent content, but no descendant media elements
Tag omission	Neither the start tag nor the end tag can be omitted
Content attributes	Global attributes, src, crossorigin, preload, autoplay, loop, muted, controls, and loading
DOM interface	HTMLAudioElement, which inherits from HTMLMediaElement

Content model and resource selection

Whether an audio element has a src attribute determines which child elements are permitted. When using child source elements, list candidates in document order and provide a type when known.

Condition	Child model	Boundary
With src	Zero or more track elements, followed by transparent content	No child source elements
Without src	Zero or more source elements, zero or more track elements, then transparent content	Transparent content cannot contain descendant media elements

<audio controls preload="metadata">
  <source src="/media/lesson.ogg" type="audio/ogg; codecs=opus">
  <source src="/media/lesson.mp3" type="audio/mpeg">
  This browser does not support the audio element.
  <a href="/media/lesson.mp3">Open the audio file</a>
</audio>
<p><a href="/transcripts/lesson.html">Read the transcript</a></p>

The type on a source is information that can help a user agent skip a candidate before fetching it. Declaring a MIME type does not establish that a container, codec, profile, server response, or network condition will allow playback. Also distinguish source selection inside picture from selection inside a media element.

Transparent content inside audio is for user agents that do not support the audio element. A browser that supports audio does not necessarily show that content when fetching or decoding a resource fails. Put a recovery link and transcript outside the element as well.

Attributes and loading boundaries

Attribute	Role in the standard	Boundary
src	URL of the media resource	When present, the child-source candidate model is not used
crossorigin	Sets the CORS mode for the media request	Read with the resource response headers and credentials conditions
preload	Author hint for expected fetching through none, metadata, or auto	Does not require the browser to honor the hint
loading	Policy for deferring off-viewport media resources	When lazy, loading deferral takes precedence over preload; support and actual timing require implementation evidence
autoplay	Requests automatic playback	User agents are not required to honor it; user preferences and playback policy can affect the result
controls	Requests user-agent playback controls	Rendering, operation, and accessibility-tree details can vary by browser and platform
muted / loop	Initial muted state / request to return to the start after playback ends	Distinguish attributes from current IDL state and actual playback permission
type	Media type hint on a child source	Not a content attribute on audio

Media state and playback requests

audio uses the same media processing model as video. The APIs below describe different states; they should not be collapsed into a single “loading percentage.”

API / concept	What it observes	Boundary
networkState	Resource selection and fetching	NETWORK_EMPTY, NETWORK_IDLE, NETWORK_LOADING, and NETWORK_NO_SOURCE
readyState	Availability of media data at the current position	HAVE_NOTHING through HAVE_ENOUGH_DATA; values need not change in sequence
currentTime / duration	Position and length on the media timeline	Finite duration cannot be assumed for unknown or unbounded resources
paused / ended	Playback processing state	Do not infer state only from the appearance of controls
play()	Playback request and a Promise for its fulfillment or rejection	It can reject with NotAllowedError when the user agent does not allow playback
Events	loadstart, loadedmetadata, canplay, playing, pause, ended, and others	Events correspond to state changes; an application cannot assume every order or timing

Changing the src or type of a child source after insertion is not a way to switch the current media resource. To explicitly change what is playing, update the media element's src and start resource selection; confirm the relationship with load() against the standard and implementation evidence.

Timed text and accessibility

A child track can specify external timed text associated with the media timeline. An audio playback area has no visual space for captions, so the presence of a track does not by itself provide a readable transcript. Separate the purpose of tracks such as chapters or metadata, the browser UI, and the actual alternatives available to users.

The 2026-09-23 Working Draft of HTML-AAM 1.0 gives audio the computed role html-audio and specifies how native controls are exposed through platform APIs. This is a specification mapping in a Working Draft. It is separate from measurements of a browser Accessibility Tree, platform API, or assistive-technology output.

WCAG 2.2 SC 1.2.1 requires an equivalent time-based media alternative for prerecorded audio-only media that falls within its scope, except when it is clearly labeled as a media alternative for text. This does not mean that the presence of an audio tag alone determines conformance. Distinguish this from caption criteria for prerecorded synchronized media and their own applicability conditions.

DOM Interface

Interface / property	Role	Boundary
HTMLAudioElement	Represents an audio element and inherits from HTMLMediaElement	Many resource-selection and playback APIs are defined on the parent interface
Audio([src])	Legacy factory function that creates an audio element	With a source, resource selection starts and preload is set to auto; this does not mean autoplay is allowed
volume / muted	Read or set playback volume and muted state	Confirm browser, platform, and user-preference constraints through implementation evidence
audioTracks	Access information about audio tracks in the resource	Do not assume multiple tracks or switching are available in every browser

Fact / Evidence

Claims from the HTML Standard, HTML-AAM, and WCAG are paired with their scope and source. The HTML-AAM entry is labeled as a Working Draft; specification mappings and implementation observations remain separate.

Evidence

1. HTML Standard: The audio element — categories, content model, attributes, and HTMLAudioElement
2. HTML Standard: Media elements — resource selection, state, playback, events, loading, and multiple tracks
3. HTML Accessibility API Mappings 1.0: audio — computed role and platform mappings; Working Draft
4. WCAG 2.2 SC 1.2.1 — prerecorded audio-only and video-only alternatives
5. Web Platform Tests: the-audio-element and media-elements — test discovery entry points. Results for nine selected files appear in Implementation Evidence and are separate from full-directory coverage.

Implementation Evidence

Normative claims are tracked separately from browser, WPT, and Accessibility Tree observations. IMPL, WPT, and AAM have partial measurements from selected scope in Chrome 154 / Windows 11 Pro build 26200.

IMPL uses a local PCM WAV fixture, WPT covers the nine files linked in the table with 28/28 assertions passing, and AAM records a Chrome Accessibility Tree snapshot. Full WPT coverage, other browsers, Windows platform APIs, and assistive technology remain unverified.

Coverage / Open Issues

• Specification reviewedMeaning, categories, context, content models with and without src, shared media attributes, and DOM interface
• Specification reviewedSource-selection, network / ready states, playback promises, and event-model entry points
• Specification reviewedHTML-AAM Working Draft and applicability of WCAG 2.2 SC 1.2.1
• Partial observationA Chrome 154 fixture confirmed source selection, media states and events, user-activated playback, and timeline. Other browsers, codecs, and error types remain open.
• Partial observationNine selected wpt.live files passed 28/28 assertions. The remaining files and nested tests in both directories were not run.
• Partial / openNative controls were observed in Chrome's Accessibility Tree. Windows platform APIs, assistive technology, transcript announcements, CORS, multiple audio tracks, and other browsers were not tested.

These partial observations are limited to Chrome 154 / Windows 11 Pro build 26200. They do not claim playback verification across browsers or codecs, complete WPT coverage, platform accessibility API or assistive-technology behavior, or that all audio content meets accessibility requirements.

Related surface

For a beginner-friendly guide to audio examples, playback controls, source candidates, and transcripts, see audio in Yugien. Parent-specific source attributes and candidate selection are covered by the source element. For the shared media state and playback model plus a visual playback area, see the video element; see the track element for timed-text kinds, attributes, and WebVTT, and picture for image-resource selection.