v0.2.1

What's Changed

Fixed

• Voice cloning silently falling back to random/auto voice when using clone:<profile> prefix (#22) — now returns HTTP 404 with a clear error message when profile is not found
• voice/speaker field in /v1/audio/speech now correctly resolves to cloned profile when a matching profile exists
• profile_id parameter in clone synthesis respects explicitly passed profile IDs
• Defensive tensor validation in script audio mixing
• Script endpoint validation and runtime failure path hardening
• Python 3.9 compatibility: replaced asyncio.timeout with asyncio.wait_for

Added

• Multi-speaker script synthesis endpoint (POST /v1/audio/script)

Full Changelog: v0.2.0...v0.2.1