Data licensing

Pronunciation data is easy to scrape and hard to trust. Ours is built.

The lexicons and language-pair mappings beneath ClearScript, available under commercial and research licences.

What's in the pack

Three layers, each reviewed by humans.

Pronunciation lexicons

Curated headword pronunciations, quality-reviewed, machine-consumable.

Respelling rule mappings

The per-pair systems that map target-language sound to first-language orthography.

Exception lexicons

The human-reviewed overrides that make the rules trustworthy at the edges.

What it's for

Products, ground truth, and research.

Learning products that need a readable pronunciation layer. TTS and alignment work that needs ground truth a human has looked at. Research that needs a respelling standard to test against. If you're unsure whether your use fits, ask — the licence terms are short and we wrote them to be read.

The data pack

Tell us what you'd build with it.

Use case and language pairs — we'll come back with what's available and what the licence looks like.

Request the data pack

Commercial and research licences available.
We reply within five working days.