Skip to content

README.md

Latest commit

 

History

History
72 lines (50 loc) · 1.86 KB

README.md

File metadata and controls

72 lines (50 loc) · 1.86 KB

MathCAT Demo TTS Proxy

Serverless proxy for speech playback. The GitHub Pages frontend posts MathCAT SSML plus a language code; the proxy selects the configured engine and returns normalized speech marks and MP3 audio.

Supported engines (set with TTS_PROVIDER):

Provider Env vars
polly (default) Uses the Lambda execution role (polly:SynthesizeSpeech)
google GOOGLE_SERVICE_ACCOUNT_JSON — full service-account JSON string
azure AZURE_SPEECH_KEY, AZURE_SPEECH_REGION

MathCAT emits W3C-style <mark name="id"/> bookmarks. Azure requires <bookmark mark="id"/>; the proxy converts that automatically when TTS_PROVIDER=azure.

API

POST /tts
Content-Type: application/json

{ "text": "<prosody>...</prosody>", "lang": "en" }

Response:

{
  "marks": [{ "time": 120, "value": "element-id" }],
  "audioBase64": "...",
  "mimeType": "audio/mpeg"
}

Deploy

cd lambda/tts
npm install
sam build
sam deploy --guided

Set TtsProvider during deploy (or in samconfig.toml). For Google/Azure, also pass the credential parameters or add them in the AWS Lambda console after deploy.

Note the TtsApiUrl output and set it in index.html:

window.MATHCAT_TTS_API = 'https://YOUR-API-ID.execute-api.us-east-1.amazonaws.com/tts';

GitHub Actions

Workflow .github/workflows/deploy-tts-lambda.yml deploys when lambda/tts/** changes.

Repository secrets for deploy:

  • AWS_ACCESS_KEY_ID
  • AWS_SECRET_ACCESS_KEY
  • AWS_REGION

Optional secrets for non-Polly engines:

  • TTS_PROVIDER
  • AZURE_SPEECH_KEY
  • AZURE_SPEECH_REGION
  • GOOGLE_SERVICE_ACCOUNT_JSON

Security

  • Credentials stay on the server; the browser only sees the proxy URL.
  • Restrict AllowedOrigins to your demo site and local dev URLs.
  • Enable billing alerts for whichever cloud TTS service you use.