Why Speech Markdown?

Here are some of the various reasons why Speech Markdown was created:

1. Shorter, simpler syntax

Instead of longer SSML:

<speak>
    A pause <break time="1s"/> then continue.
</speak>

Use more concise Speech Markdown:

A pause [1s] then continue.

Use a plain text formatter and convert Speech Markdown:

This is (an important)[emphasis] announcement
Are you ++listening?++

to plain text:

This is an important announcement
Are you listening?

Use platform-specific formatters to convert Speech Markdown:

Are you ++listening?++

to SSML for a given platform:

<speak>
    Are you <emphasis level="strong">listening?</emphasis>
</speak>

<speak>
    Are you <emphasis level="strong">listening?</emphasis>
</speak>

In addition to formatters for:

Formatters can be created for other platforms:

A platform’s implementation of W3C standard SSML may not be complete and could differ from another platform.

One such example is the SSML <phoneme> tag that Amazon Alexa supports and Google Assistant does not.

When you use the ipa modifier in Speech Markdown:

You say, (pecan)[ipa:"pɪˈkɑːn"].
I say, (pecan)[ipa:"ˈpi.kæn"].

The formatter will render in Amazon Alexa SSML:

<speak>
    You say, <phoneme alphabet="ipa" ph="pɪˈkɑːn">pecan</phoneme>.
    I say, <phoneme alphabet="ipa" ph="ˈpi.kæn">pecan</phoneme>.
</speak>

But the formatter will ignore it in Google Assistant SSML:

<speak>
    You say, pecan.
    I say, pecan.
</speak>

Some Speech Markdown:

(I am not a real human.)[whisper]

will be converted to a specific flavor of SSML using platform-specific formatters:

<speak>
    <amazon:effect name="whispered">I am not a real human.</amazon:effect>.
</speak>

<speak>
    I am not a real human.
</speak>

Formatters can be configured to choose a similar substitution for platform-specific elements.

Some Speech Markdown:

(I am not a real human.)[whisper]

will be converted to a specific flavor of SSML using platform-specific formatters:

<speak>
    <amazon:effect name="whispered">I am not a real human.</amazon:effect>.
</speak>

and pick a close substitution if a given tag is not supported:

<speak>
    <prosody volume="x-soft" rate="slow">I am not a real human.</prosody>
</speak>

When using multiple SSML tags on a single text block, the modified text can get lost in the nesting:

<speak>
    My favorite chemical element is <prosody volume="x-loud" rate="slow" pitch="low"><sub alias="aluminum">Al</sub></prosody>.
</speak>

Speech Markdown keeps the modified text together with the modifiers after:

My favorite chemical element is (Al)[sub:"aluminum";volume:"x-loud";rate:"slow";pitch:"low"]