chore: version packages

[github-actions]

This PR was opened by the Changesets release GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.

Releases

spoken-text@0.2.0

Minor Changes

• e49327f: Read a whole document, not just a passage.

  • <SpokenText> accepts elements as well as a string. It walks the tree,
    wraps the words and leaves the structure alone: an h2 stays an h2, a link
    stays a link.
  • Each heading, paragraph, list item and blockquote is its own alignment
    request. The first is fetched on mount, the next is warmed while the current
    one plays, and clicking a word in a block nobody has asked for yet fetches it
    and plays from there. currentWordIndex is one index across the document, and
    segments says where each block begins and ends.
  • duration counts unloaded blocks at a reading pace until their audio lands,
    and durationIsEstimate says when it is doing so. <Player> shows the total
    dimmed, with a tilde, while it is a guess.
  • <SpokenTextProvider> holds the controller, so <Player> can live in a
    sticky header rather than next to the text. useSpokenTextController() reads
    it, for building a player of your own.
  • skip and only decide what is spoken, taking a tag, a .class, an
    [attr], or a predicate over the React element. data-spoken and
    data-spoken-skip are the per-element forms. Skipped content still renders
    where it was written and is dropped only from the text handed to the aligner,
    so the words either side of an inline <code> stay correctly timed.

  Breaking: <Transport> is now <Player>, and TransportProps /
  TransportClassNames are PlayerProps / PlayerClassNames. Same component,
  better name; there is no alias.

• e13bf4e: The highlight reads as one band rather than a row of tiles. The whitespace
  between two words is now a span of its own, carrying data-spoken-state and
  the past colour once the word after it has been reached, and no word box is
  rounded, padded or overlapped on the way. Only the word being spoken is a box
  of its own. classNames.separator restyles the gaps.

• e7ba6c2: Say what is not on the page.

  • say takes a rule per selector and puts words into the audio that are not in
    the markup:

    <SpokenText
      say={{
        h2: ({ count }) => ({ before: `Section ${count}.` }),
        pre: () => "A code sample, skipped.",
        svg: ({ element }) => `A drawing. ${element.props["aria-label"]}`,
      }}
    >
      …
    </SpokenText>

  • { before } and { after } keep the element's own words and speak around
    them; a plain string replaces them. Keys select the way skip does, and the
    first one that matches an element wins.

  • The words a rule adds are hidden words: they are in the word list and in the
    text handed to the aligner, at the point the element sits, so everything
    around them stays exactly timed — and nothing is rendered for them.
    DisplayWord.hidden marks them, no word on the page is current while one is
    spoken, and the highlight band waits for the voice to come back to the page.

• ee1d15b: Speech that brings its own timings, and a kind for every block.

  • elevenlabsSpeech is a new adapter. One call returns the audio and a
    timestamp for every character, which the adapter groups into words, so there
    is no transcription step and no drift between what was said and what was
    heard. Plain fetch, no SDK, no new dependency. Needs ELEVENLABS_API_KEY
    (or ELEVEN_LABS_API_KEY); defaults to George, a premade voice that works on
    every plan.
  • SpeechAudio can carry words and duration, and transcribe is now
    optional. Supply neither and the handler says so, naming both ways out.
  • Every block carries a kind — "heading", "paragraph", "list" or
    "quote" — from the tag it was written as. The client sends
    { content, kind }, the handler passes it to speech and to hash, and
    segments[].kind reports it. openaiSpeech takes instructions as a
    function of the kind; elevenlabsSpeech takes voiceSettings the same way.
  • FetchAlignment is now (text, { kind }) => Promise<Alignment> and hash is
    now (text, kind) => string. An adapter written as (text) => … still fits;
    a hash you wrote yourself keeps working and simply ignores the kind.

  Cache note: the default hash covers the kind as well as the text, so
  existing entries are missed once and regenerated.

• e13bf4e: Options written on a <SpokenText> inside a <SpokenTextProvider> are honoured
  rather than ignored. The provider's own options are the defaults, so a
  debounceMs can sit next to the text it is about.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
spoken-text	Ready	Preview	Sep 3, 2026 3:58am UTC