with one click
elevenlabs-tts
// Generate high-quality speech for Pompom companion using ElevenLabs TTS API v3
// Generate high-quality speech for Pompom companion using ElevenLabs TTS API v3
| name | elevenlabs-tts |
| description | Generate high-quality speech for Pompom companion using ElevenLabs TTS API v3 |
Generate high-quality speech for Pompom companion using ElevenLabs Text-to-Speech API v3.
pompom-voice.ts ElevenLabsEngine)/pompom:voice test)g6xIsTj2HwM6VR4iXFCweleven_v3ELEVENLABS_API_KEY from environment (~/.env.secrets)enable_logging=false query param + xi-no-log: true header| Name | Voice ID | Style |
|---|---|---|
| Jessica Anne Bogart - Chatty (default) | g6xIsTj2HwM6VR4iXFCw | Chatty, friendly |
| Lily - Soft, Cute and Sweet | Pt5YrLNyu6d2s3s4CVMg | Soft, sweet |
| Cherry Twinkle - Bubbly and Sweet | XJ2fW4ybq7HouelYYGcL | Bubbly |
| Flicker - Cheerful Fairy | piI8Kku0DcvcL6TTSeQt | Cheerful |
| Jessica Anne Bogart - Eloquent | flHkNRp1BlvT73UL6gyz | Eloquent, dramatic |
| Sarah - Mature, Confident | EXAVITQu4vr4xnSDxMaL | Professional |
The voice engine is in extensions/pompom-voice.ts (ElevenLabsEngine class). Speech events flow:
say() in pompom.ts fires a SpeechEventenqueueSpeech()Eleven v3 supports emotional audio tags โ used throughout Pompom's speech:
[excited] โ energetic reactions (ball, dance, game, treats)[happy] โ warm moments (petting, hugs, feeding, accessories)[whispers] โ quiet moments (sleep, nap)[sighs] โ contentment (waking from nap)[mischievously] โ playful moments (peekaboo)[sings] โ musical moments (la la la)[curious] โ weather observations (clouds rolling in)[concerned] โ weather warnings (storm brewing)Rules:
[excited] Catch the stars!sanitizeSpeechText()v3 does NOT support SSML <break> tags. Use instead:
Audio tags like [laughs], [sighs], [excited] are ElevenLabs v3 only.
When Kokoro or Deepgram is the active engine, tags are automatically stripped
before synthesis โ the text "Hello!" still reaches the engine, but "[laughs]" is removed.
No engine loses features because of another engine's limitations.
Kokoro supports Markdown pronunciation: [word](/IPA/) and stress: [word](+2)
Deepgram Aura-2 uses natural prosody from punctuation โ no markup needed.