LingoLoop Allosaurus service

This companion service performs experimental acoustic pronunciation assessment. It converts uploaded microphone audio to mono WAV, recognizes phones with Allosaurus, generates reference phones with eSpeak NG, aligns the two phone sequences, and returns accuracy, completeness, fluency, and phone-level hints.

Run with Docker

docker build -t lingoloop-allosaurus .
docker run --rm -p 7860:7860 lingoloop-allosaurus

Set the website runtime secret ALLOSAURUS_SERVICE_URL to the public HTTPS URL of this service.

Run on Render Free

Create a Web Service from this directory's public Git repository and apply the included render.yaml Blueprint. Keep the payment method empty so exceeding a free usage allowance suspends the service instead of creating a charge. Free services sleep after 15 minutes without traffic, so the first pronunciation request can take longer.

Set the same random value as ALLOSAURUS_SHARED_SECRET on the service and ALLOSAURUS_SERVICE_KEY on the website so the assessment endpoint cannot be used anonymously outside LingoLoop.

The universal model is experimental for all languages other than its dedicated English model. Scores must not be presented as clinical or certification-grade assessment.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support