mirror of
https://github.com/calesthio/OpenMontage.git
synced 2026-08-22 16:23:22 +08:00
Neural TTS via the synchronous REST v1 endpoint (SSML body, no token exchange or job polling). Shares one Speech resource with azure_stt — AZURE_SPEECH_KEY + AZURE_SPEECH_REGION unlock both directions; optional AZURE_TTS_ENDPOINT overrides the TTS host (a different subdomain than the STT endpoint). piper_tts remains the default offline path. - tools/audio/azure_tts.py: azure_tts tool (capability=tts), voice shortlist aliases, SSML prosody/style, mp3/wav output, cost tracking - tests/tools/test_azure_tts.py: contract, discovery, status, SSML, and mocked execute tests (21 tests, no live network) - .agents/.claude skills: azure-text-to-speech Layer-3 skill - docs: PROVIDERS.md section + tables, ARCHITECTURE.md inventories, AGENT_GUIDE.md + skills/INDEX.md rows, asset-director TTS cheatsheet, .env.example