ElevenLabs TTS

Use this skill when implementing or debugging ElevenLabs text-to-speech in production code. It emphasizes architecture decisions first, then API usage.

Philosophy: Speech Is a Product Surface

TTS is not just an API call; it is a UX contract across identity, latency, intelligibility, and reliability.

Before implementing, ask:

• Is this interaction realtime, near-realtime, or offline pre-generation?
• What matters most here: naturalness, speed, cost, or deterministic reproducibility?
• Where is the trust boundary, and how are API credentials protected?
• What fallback should happen if voice generation fails or times out?

Core principles:

1. Delivery-first design: Choose pipeline and endpoint from latency/quality targets, not preference.
2. Secrets never in clients: API keys belong server-side; clients get short-lived scoped tokens when needed.
3. Deterministic contracts: Standardize request and response shapes so retries and caching are safe.
4. Graceful degradation: Always define timeout, retry, and fallback behavior before shipping.