✨ Features

Eleven3 is a private production bench for turning text, dialogue, sound prompts, and voice-design ideas into downloadable audio without exposing provider credentials to the browser.

🧩 Workspace

• Speech: single-voice Eleven v3 TTS with expressive tag support and keyboard generation.
• Dialogue: multi-speaker scripts with role assignment and per-line voice routing.
• Voices: searchable ElevenLabs library with automatic default selection.
• Designer: text-to-voice previews, loudness, guidance, reference audio, and optional metadata enrichment.
• SFX: sound-generation prompts with duration, prompt influence, and loop-friendly presets.
• Batch: repeatable queued rendering for prepared prompt sets.
• History: generated audio review, replay, and download from local session state.

🎚️ Audio Flow

• choose voice and model.
• write ear-first text with sparse Eleven v3 tags.
• send generation through the app proxy.
• play from the persistent bottom player.
• download individual results or batch outputs when available.

🧠 Defaults

• speech model: eleven_v3.
• voice design model: eleven_ttv_v3.
• default output: mp3_44100_128.
• supported variants: higher-bitrate MP3, PCM, μ-law, and designer WAV formats.

🛡️ Guards

• login required before workspace access.
• /api/elevenlabs routes rate-limit per auth cookie.
• browser never receives ElevenLabs credentials.
• voice list recovers when downstream actions clear state.
• blob URLs are cleaned up before replacement.

---