Benchmarks

Coming Soon

Automated, reproducible benchmarks for TTS and STT models. Run standardized test suites directly in your browser and compare results across models.

What to Expect

Harvard sentences for TTS, LibriSpeech clips for STT. Curated datasets that test pronunciation, prosody, and accuracy across a wide range of inputs.

Same hardware, same browser, consistent results. Every benchmark run is deterministic and comparable, so you can trust the numbers.

Word Error Rate (WER), Real-Time Factor (RTF), MOS predictions, and more. Get objective scores without manual evaluation.

See how models rank across different benchmarks. Compare performance on specific tasks and find the best model for your use case.

Subscribe to be the first to know when benchmarks launch.