Glossary · Speech
What is SSML in text to speech?
SSML
SSML, the Speech Synthesis Markup Language, is a W3C standard for adding instructions to text before a voice reads it. Tags can ask for a pause, stress a word, or say how a number or date should be read.
On a call
An IVR prompt in SSML might read: <speak>आपका OTP है <say-as interpret-as="characters">4821</say-as><break time="500ms"/> धन्यवाद</speak>. The tags ask for the code one digit at a time, then a half-second pause.
AI agents from ₹1 a minute.¹
Explained
Here is how it works, in plain words.
SSML grew up with IVR systems. A menu prompt is written once and played a million times, so it pays to control every pause and every number in it. The W3C published the first version in 2004. Version 1.1 followed in 2010.
The markup wraps text in a <speak> element and adds tags inside it. <break> inserts a pause. <emphasis> stresses a word. <say-as> tells the engine what kind of thing it’s reading, such as a date or a string of characters. <phoneme> gives an exact pronunciation, and <prosody> changes speed, pitch or volume. There’s also <lang>, for switching language mid-sentence.
How much of that an engine honours is up to the engine. Most cloud TTS services support a subset, and the same tag can sound different from one provider to the next. Test the tags you rely on, in the voice you’ll ship.
For voice agents, SSML matters less than it used to. The reply is written by a language model a moment before it’s spoken, so the model can write “पाँच हज़ार रुपये” instead of a figure, and keep sentences short enough to need no extra pauses. Newer neural voices also place pauses and stress from the words themselves.
SSML still earns its keep in fixed text, like a legal notice or an OTP prompt. Where it’s missing, plain-text habits do most of the same work. A comma gives the voice a breath, and words read more reliably than digits.
One caution with mixed scripts: markup inside Hinglish text gets fiddly fast, so wrap whole phrases in a tag where you can.
In Voice Infra
This is what Voice Infra does about it.
Mira TTS takes plain text. SSML isn't among the inputs our docs list, which are the text, the voice, the format and the sample rate. To change how something is read, write it the way it should sound, the same way our prompting guide writes amounts and identifiers.
Questions
People often ask us these questions.
Does Mira TTS support SSML?
Our docs don't list it. The endpoint takes plain text, so write the words the way they should be spoken. If SSML matters for your project, ask us before you build on it.
Do I need SSML for a voice agent?
Usually not. The language model writes the reply, so it can write numbers as words and keep sentences short. SSML earns its place in fixed prompts, like IVR menus, that people write by hand.
Still deciding? Ask us.
Fifteen minutes with the team that builds it. Bring your script, or your hardest call.
Book 15 minutes