✨ Features

• shell:
  • SvelteKit application using Svelte 5 runes and Skeleton UI.
  • sticky full-width navigation with six active work areas.
  • contextual right panel for voices, models, output formats, examples, and mode-specific settings.
  • full-width History mode without the right sidebar.
• work_areas:
  • Speech:
    • single-voice text-to-speech.
    • dialogue mode folded into the Speech tab.
    • Ctrl or Cmd plus Enter generation shortcut.
    • multilingual prompt examples with expressive Eleven v3 tags.
  • Voices:
    • searchable voice library loaded from ElevenLabs.
    • automatic default voice selection when no voice is chosen.
    • voice parsing helper keeps names, taglines, and descriptions consistent.
  • Designer:
    • text-to-voice design previews.
    • loudness, guidance scale, output format, prompt strength, and optional reference audio.
    • save path can enrich voice metadata through OpenRouter when the optional key is configured.
  • SFX:
    • sound-generation prompt workflow.
    • duration, prompt influence, and loop-friendly ambient presets.
    • quick examples for explosions, rain, cars, bells, fire, wind, and other common effects.
  • Batch:
    • queued generation items with processing state and per-item results.
    • designed for repeatable prompt sets rather than one-off manual generation.
  • History:
    • local history capped by application state.
    • tracks tts, dialogue, sfx, designer, and batch sources.
    • stores generated audio as data URLs for later playback and downloads.
• audio_experience:
  • bottom player persists across tabs.
  • blob URLs are cleaned up before replacement.
  • current voice name and source type are shown alongside playback state.
• model_defaults:
  • speech model: eleven_v3.
  • voice design model: eleven_ttv_v3.
  • default output: mp3_44100_128.
  • alternate outputs include higher bitrate MP3, PCM, μ-law, and designer WAV formats.
• reliability_guards:
  • app requires login before workspace access.
  • /api/elevenlabs routes are rate-limited per auth cookie.
  • voice list auto-recovers when cleared by downstream actions.
  • API calls are server-side proxies so browser code never receives ElevenLabs credentials.