First I want to share how excited I am that there's a CSS Speech module. Considering that language originates in the verbal form and that most writing systems are based on the verbal form, I think it's long-overdue that we address how content can be consumed this way. Tone and stress are significant aspects of communication and think this module will help ensure that users don't experience a loss of information.

With that consideration, I would like to suggest / request two things:

Problem: Information loss in spoken modality due to lack of stress and tone

Solution: We allow the default visual distinctions of certain phrasing elements inform some default auditory distinctions:

• em : voice-stress: moderate
• strong : voice-stress: strong
• mark : voice-stress: strong; voice-pitch: high
• small, sub, sup : voice-stress: reduced
• i, q: voice-balance: leftwards; pause: weak;
• i : voice-rate: slow (an unbalanced slow voice with a pause could reflect alternate thoughts and transliterations)
• dfn, dt: voice-rate: slow; pause: weak;
• del,s,strike: voice-range: low; voice-volume: x-soft (less range and much softer, indicating contrast to regular)
• ins: voice-range: medium; voice-volume: soft `
• u: voice-balance: rightwards; ¿¿speak-as: spell-out??; voice-rate: slow; (unbalance slow voice indicating it's annotated in someway. especially if spelled out)
• pre, code, var, samp, data: voice-range: x-low (monotone could be similar to monospace)
• kbd, data: speak-as: literal-punctuation; (guarantees that !.; are read aloud)
• var, samp: voice-pitch: low
• var: pause: weak
• samp: voice-stress: moderate
• abbr: speak-as: spell-out

Problem: We don't have a way define default and alternate voice styles

Certain elements carry semantics that communicate, "not the narrator's original words / phrasing" such as:

• q
• blockquote
• samp

We recognize quoted text in narrations because the narrator has altered their voice. Whether they read the text higher, lower, more feminine or masculine, or with a different accent. The narrator alters their dictation to indicate, "this is not the narrator's voice."

While we have all of the individual parts for a "narrator voice" and "non-narrator voice", we don't have a way to point the whole collection of attributes together so that we can apply them.

Solution: A @voice-character at-rule and/or reading-style property

I could see something like how we use @keyframes and animation to define a set of attributes and apply them to an element:

@voice-character Narrator {
  voice-family: young neutral;
  voice-stress: normal;
  voice-pitch: medium;
  voice-range: medium;
}

@voice-character Excerpt {
  voice-family: old neutral;
  voice-stress: moderate;
  voice-pitch: high;
  voice-balance: leftwards;
}

@voice-character Code {
  voice-family: young neutral;
  voice-stress: none;
  voice-range: x-low;
}

body {
  reading-style: Narrator;
}

q, blockquote {
  reading-style: Excerpt;
}

pre, code, var, samp, data {
  reading-style: Code;
}