Provider-specific SSML and plain-text preparation

SSML examples: mark up a welcome, keep the spoken words intact.

Keep the message. Choose the input format deliberately.

A useful SSML example lets you see both the instructions and the words the listener should hear. This guide pairs one complete Azure-specific XML document with its exact plain-text reading. Follow an original exhibition welcome from a finished brief to a deliberate pause, then choose the input edition for the service you actually use.

Six parts of this Azure SSML example
Element or attributeDocumented roleChoice in this original example
speak with version="1.0"Required root and SSML version.Wrap the entire welcome once.
xmlns and xml:langMarkup namespace and root language.Use the SSML namespace and en-US.
voice with nameAzure voice selection by name.Use the documented en-US-JennyNeural identifier.
pParagraph grouping.Separate the exhibition introduction from the visitor invitation.
sSentence grouping.Keep each complete sentence in its own element.
break with time="500ms"Explicit break with a millisecond time value.Put the author-selected break before the touch-sample explanation.

Choose an Azure SSML example for a finished message

Azure’s SSML reference documents XML structure, paragraph and sentence grouping, and explicit breaks. Supported tags vary by voice type.

The writing brief is an invented exhibition for adults called Thread & Light. Two woven panels hang in different places: blue near the entrance and orange on the back wall. A folded cloth on a low table is a separate touch sample. The welcome should distinguish those objects before inviting the visitor to choose a panel. None of the requested visitor actions has happened yet.

Finish that message before decorating it. The accepted script gives the location of each panel, explains what may be handled and offers a next step without choosing it for the listener. It contains no stage directions, background-music request or claim that an exhibition recording already exists.

A complete original XML welcome and its exact spoken text

The XML sample uses one voice throughout. Microsoft lists en-US-JennyNeural in its English United States voice table. Confirm the voice available to your Azure setup before using the document; this guide has not synthesized it.

The complete XML and plain-text editions below contain the same eight sentences in the same order. The separation between the two plain-text paragraphs is a reading layout. The XML adds one deliberate production request after the sentence about cotton thread, before the explanation of the touch sample.

Treat the exhibition name, locations and invitation as the accepted writing. Keep the code heading, the plain-text heading and this commentary outside the spoken input. The listener should hear a welcome, rather than a tutorial about how the welcome was marked up.

Original Azure SSML exhibition welcome
<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xml:lang="en-US">
  <voice name="en-US-JennyNeural">
    <p>
      <s>Welcome to Thread &amp; Light, an exhibition of woven colour for adults.</s>
      <s>The blue panel hangs near the entrance.</s>
      <s>The orange panel hangs on the back wall.</s>
      <s>Both were woven from the same bundle of cotton thread.</s>
      <break time="500ms" />
    </p>
    <p>
      <s>The folded cloth on the low table is a separate touch sample.</s>
      <s>You may handle that sample; please leave the hanging panels in place.</s>
      <s>Begin with either panel, then look for one colour that appears in both.</s>
      <s>When you are ready, the labels beside each panel offer the maker's notes.</s>
    </p>
  </voice>
</speak>

Exact plain spoken text and a tag-by-tag review

Welcome to Thread & Light, an exhibition of woven colour for adults. The blue panel hangs near the entrance. The orange panel hangs on the back wall. Both were woven from the same bundle of cotton thread.

The folded cloth on the low table is a separate touch sample. You may handle that sample; please leave the hanging panels in place. Begin with either panel, then look for one colour that appears in both. When you are ready, the labels beside each panel offer the maker's notes.

The table maps each component to a specific writing choice. Start at the outside of the sample, then read inward to the sentences. This makes it easier to distinguish a setting that applies to the document from the words naming a real object within the fictional exhibition.

Only the two content groups differ in purpose: the first establishes the panels, and the second addresses the visitor. The grouping does not add an unseen speaker or a new exhibit. Keep the voice identifier out of the welcome itself; the exhibition has no character named Jenny.

Use the smallest set of instructions that expresses your intention. This sample needs neither pronunciation spellings nor pitch, emotion, rate or volume adjustments. If another passage needs a different capability, check that provider and voice rather than adding an unfamiliar tag from an unrelated example.

Make one pause decision without promising a recorded result

The chosen boundary finishes the introduction before discussing what the visitor may touch. The author-selected 500ms value is a request to assess, not a measured best pause or an audio result. A longer gap inside “cotton thread” would divide the material name rather than separate the two ideas.

If you later compare a version with a different break value, keep the eight sentences fixed so you know which deliberate decision changed. Record the value you requested separately from any observations you actually make. Leave observations blank until you have a result to review.

The sample says to begin with either panel. Preserve that choice during review. Changing a pause cannot justify rewriting the instruction as if the visitor had already approached the blue panel or handled the cloth. Wording, production instructions and listener actions remain different kinds of information.

Prepare the plain-text edition without losing or adding words

In this example, the name’s & becomes &amp; inside XML text, following Microsoft’s documented entity notation.

For this example, use the supplied plain-text edition rather than copying a mixture of words and code. Read it against the eight source sentences in order: welcome, blue location, orange location, shared thread, touch sample, handling instruction, panel choice and labels. Both hanging panels remain in place, and the folded sample stays a separate object.

Keep the 500ms decision in a production note when moving to a different input path. A plain-text edition preserves the words, not a promise that another service will interpret the same markup or reproduce identical delivery. Review the supported controls in the destination separately.

Use Honeyline’s plain-text narration path separately

For Honeyline Make audio, use the plain spoken edition rather than the SSML document. Type or paste the complete reviewed words into the current script editor. Use its offered narrator, title and cue controls, then check current passage and project validation, any sign-in requirement and the displayed quote before generating.

If you want an explicit pause in that workflow, use the supported Pause control and inspect the cue beside the surrounding words. The separate text-to-speech pauses guide explains that control. Do not treat the Azure break tag as a Honeyline command or assume that copying the XML transfers its settings.

Keep two clearly labelled working copies if you need both formats: one Azure SSML document and one plain spoken script. Make the same wording correction in both, then compare their spoken words again. This guide supplies writing and markup examples; it contains no generated audio, playback evaluation or measured running time.

Official Azure SSML references

Microsoft Learn references checked November 2, 2026 UTC. The exhibition welcome and its markup arrangement are original. Documentation review did not generate or test audio.

Frequently asked questions

What is the SSML example on this page for?

This is an Azure-specific markup example for the fictional exhibition welcome. Its complete plain-text edition serves a separate narration-input workflow.

Why include a complete plain-text version?

It makes every intended spoken word visible without the production instructions. Here the XML and plain-text copies preserve all eight sentences, their order and punctuation. It also gives you a clean edition for a narration path that expects spoken text.

Can I use the Azure voice name as a Honeyline narrator?

Choose among the voices actually offered by the service you are using. Microsoft documents the Azure voice identifier used in this example; it is not a promise that Honeyline offers that voice or imports an Azure voice selection.

Does the 500ms break prove the welcome has the right timing?

No audio has been generated or reviewed for this example. The break is one deliberate starting request between two ideas. Assess any actual result with its surrounding speech rather than treating a written value as a listening verdict or total-duration guarantee.

Should I paste the SSML into Honeyline Make audio?

Use the plain spoken edition for the Honeyline workflow described here. Keep Azure XML and production notes outside that input, and use the currently supported editor controls for cues. This guide does not present SSML parsing or import as a Honeyline capability.

How should I revise the welcome later?

First decide which spoken sentence needs to change and why. Update that sentence in both working editions without adding a visitor action that has not happened. Check the complete word sequence again, then review each destination’s production settings separately.

The words travel separately from provider markup.

Bring the reviewed plain spoken text to Make audio, choose an available narrator and check current controls, validation, access requirements and the displayed quote.