SSML (Speech Synthesis Markup Language) lets you control how text is spoken: where to pause, which words to stress, how fast, how high and how loud. On ttsMP3 you can type SSML tags directly into your text – no setup needed. All examples below work with the 60 regular voices (Amazon Polly); the AI voices support pauses. Try them in the free converter.
Copy any example into the converter. You do not need to add <speak> – ttsMP3 adds it automatically.
| Effect | Example | Notes |
|---|---|---|
| Pause | <break time="1s"/> | Up to 10 seconds with regular voices; also strength="strong". AI voices: whole seconds from 1 to 60. |
| Emphasis | <emphasis level="strong">really</emphasis> | Levels: strong, moderate, reduced. |
| Speed | <prosody rate="slow">…</prosody> | x-slow to x-fast or a percentage like 80%. |
| Pitch | <prosody pitch="high">…</prosody> | x-low to x-high or relative like -20%. |
| Volume | <prosody volume="loud">…</prosody> | silent, soft, medium, loud, x-loud or dB like +6dB. |
| Whisper | <amazon:effect name="whispered">…</amazon:effect> | Amazon Polly specific. |
| Spell out | <say-as interpret-as="characters">ABC</say-as> | Reads letter by letter. |
| Numbers and dates | <say-as interpret-as="date" format="dmy">28-10-2026</say-as> | Also cardinal, ordinal, digits, telephone, time. |
| Abbreviation | <sub alias="World Wide Web Consortium">W3C</sub> | Speaks the alias instead of the written text. |
| Pronunciation | <phoneme alphabet="ipa" ph="pɪˈkɑːn">pecan</phoneme> | Exact pronunciation in IPA. |
| Other language | <lang xml:lang="fr-FR">Je ne sais quoi</lang> | Pronounces a phrase in another language. |
| Paragraph / sentence | <p>…</p> <s>…</s> | Adds the natural pause of a paragraph or sentence. |
The most used tag. Put it where the voice should stop for a moment:
Welcome to our course. <break time="2s"/> Let us start with chapter one.
With the regular voices a pause can be up to 10 seconds; for longer silences combine several breaks. With the AI voices pauses from 1 to 60 whole seconds are possible, for example <break time="5s"/>. A full stop and a space between sentences also improves the rhythm.
Stress a word: I <emphasis level="strong">really</emphasis> like that.
Slow down an important sentence: <prosody rate="slow">Please read the safety instructions.</prosody>
Combine attributes: <prosody rate="90%" pitch="-10%" volume="loud">Attention, please.</prosody>
<sub alias="Doctor">Dr.</sub> SmithYour code is <say-as interpret-as="characters">X7K2</say-as>.<say-as interpret-as="telephone">+1 555 0100</say-as><phoneme alphabet="ipa" ph="…">Name</phoneme> with the IPA spelling.<lang xml:lang="de-DE">Gesundheit</lang>ttsMP3 adds its own simple syntax for conversations: start each line with the speaker name of a regular voice.
[speaker:Brian] Hello Emma[speaker:Emma] Hey Brian, how are you?
Remove accents from names (Léa → Lea). Listen to all speakers on the voice samples page.
<prosody>…</prosody>; single tags end with />, like <break time="1s"/>.& instead of & and < instead of <, or simply write "and" / "less than"."1s", not typographic quotes “1s”, which some editors insert automatically.SSML (Speech Synthesis Markup Language) is a set of tags that control how a text-to-speech voice speaks: pauses, emphasis, speed, pitch, volume and pronunciation.
Insert <break time="1s"/> where the pause should be. With ttsMP3 regular voices pauses can be up to 10 seconds, with the AI voices from 1 to 60 seconds.
No. On ttsMP3 you type the tags directly into your text; the <speak> wrapper is added automatically.
With the regular voices (Amazon Polly) all common Polly SSML tags such as break, emphasis, prosody, say-as, sub, phoneme, lang and amazon:effect. The AI voices support break.
Usually a tag is not closed, misspelled, or the text contains a bare & or < character. Close every tag and write & or < instead.
Related: Amazon Polly voices online · Text to speech for YouTube voiceovers · Commercial use of text to speech · Text to MP3 · Voice samples · Presentations and e-learning · SSML guide · FAQ · API