ElevenLabs · Contract tests passed · community-built lab
ElevenLabs API Studio
Inspect normalized model and voice catalogs, then deliberately run bounded Text to Speech or prerecorded Speech to Text requests without exposing a permanent API key to the browser.
Converged safety boundary
This is an independent ONE Voice Lab integration, not an official ElevenLabs product, benchmark, endorsement, or production-readiness claim.
Deterministic fixtures verify the canonical adapters and account-scoped normalization boundary. Live execution stays disabled unless current server policy, cost admission, configuration, and health state all permit it; no live ElevenLabs request was used to verify this convergence.
Server boundary
ElevenLabs configuration
The permanent API key stays on the server. This page receives only normalized catalogs, audio bytes, transcript results, and safe request metadata.
- Credential readiness
- Server key not configured
- Execution policy
- Canonical execution policy disabled
No request runs on page load. Catalogs require a deliberate click; TTS and STT require an additional human confirmation. Server-instance request limits and provider kill switches remain active.
1. Load safe catalogs
Loads model metadata and the first bounded page of voices. Account internals, preview URLs, and the server credential are not serialized.
Catalog requests have not run.
Explicit provider usage
2. Text to Speech
84 / 1,000 characters
No synthesis request has run.
Explicit provider usage
3. Prerecorded Speech to Text
Maximum 10 MB. The file is forwarded for this request and is not persisted by the Lab route.
No transcription request has run.
Evidence and safety boundary
- Repository verified: fixed upstream routes, bounded inputs, server-only credential access, normalized responses, and mocked route tests.
- Provider documentation verified: the listed REST contracts are grounded in current public ElevenLabs API documentation.
- Manual verification required: this account's entitlement, live output, latency, quality, quota behavior, and production suitability.
- Not implemented: realtime STT, streaming TTS, voice cloning, agents, arbitrary API proxying, and automatic execution.