Can I build my own Eleven Labs with AI?

Build a simple AI voice studio that uses a text-to-speech API (do not train a model).

**Objective:** Let a creator paste a script, choose a voice, generate audio via an API, and download or save the result.

**Target user:** A content creator or product builder who wants a custom voiceover workflow without rebuilding ElevenLabs.

**Core functionality:**
- Paste or write a script
- Select a voice from a configured list (API voice IDs)
- Generate audio via TTS API
- Play and download MP3/WAV
- Save generations to a personal library
- Show estimated character usage

**Pages/screens:**
- Login
- Studio (script + voice + generate)
- Library of past generations
- Settings (API key stored server-side, default voice)

**User flows:**
1. User opens Studio and pastes a script
2. User picks a voice and clicks Generate
3. App calls TTS API, stores file, shows player
4. User downloads audio or reopens it from Library

**UI direction:** Minimal creator tool, large script textarea, clear Generate CTA, audio player under the result. No fake "we trained our own model" branding.

**Data model:**
- User: { id, email }
- Generation: { id, userId, script, voiceId, audioUrl, charCount, createdAt }

**Authentication:** Required so libraries stay private.

**APIs/integrations:** ElevenLabs API (or OpenAI TTS / similar). Store the API key on the server, never in the browser. Object storage for audio files.

**Responsive requirements:** Studio usable on tablet; desktop preferred for long scripts.

**Error states:** API failures with readable errors; character limit warnings; empty library state.

**Explicit exclusions:** No custom model training, no voice marketplace, no real-time conversational agents, no dubbing pipeline in v1.

**Suggested implementation order:**
1. Auth + settings for server-side API key
2. Studio form → TTS API → file storage
3. Audio player + download
4. Library history
5. Usage counters and polish