Text to Speech page: Speak Text section with voice menu and markup guide

[dsward2]

What

A new Speak Text section at the bottom of the web UI's Text to Speech page (Devices):

• A text box, prefilled with <speak>It is four oh nine <break time="700ms"/> <prosody rate="80%">on AntennaHead Radio.</prosody></speak>.
• A Voice menu of the installed voices. "Default voice" means the Text to Speech voice chosen in Configuration.
• A Speak button that speaks the text once through the live pipeline (PCMSpeechSynth → sox → PCMUDPSender) and starts the web player, the same way Listen does. When it finishes, the filler starts, as it does after a one-pass Listen.
• Below that, a collapsible guide, "Controlling the voice: pauses, speed, pronunciation", covering the two markup methods:
  • SSML for modern voices, with a table of which tags actually changed the audio when I rendered them with the Premium Ava voice. prosody rate and volume, break, say-as characters and phoneme work; prosody pitch barely changes anything; emphasis and sub are ignored.
  • [[slnc]], [[rate]], [[pbas]] and [[pmod]] embedded commands for classic voices.
  • A warning not to mix the two.

How

• Text that starts with <speak is sent with --ssml. That's the same rule StationDirector's announcer uses, and the rule is needed. In tests with file input:
  • SSML with --ssml: 3.84 s, with the pause and the slower ending.
  • Plain text with --ssml: no audio at all.
  • SSML without --ssml: 10.8 s, because the tags are read aloud.
• SDRController: the launch half of startTextToSpeech moved into startSpeechSynthPipeline(text:repeatForever:voiceIdentifier:ssml:dying:), which the new speakText(_:voiceIdentifier:) shares. makeSpeechSynthTaskItem accepts an optional voice and SSML flag. Existing folder behavior is unchanged: same voice setting, no --ssml.
• New route: POST /speaktextbuttonclicked.html with body {text, voice}.
• antennahead.js: speakTextButtonClicked(form).

Verification

• Debug build succeeds, and node --check passes on antennahead.js.
• I tested PCMSpeechSynth's file input with and without --ssml as described above.
• Not yet tested in the running app. That needs the installed AntennaHead restarted, which interrupts the live stream.

🤖 Generated with Claude Code