Measured release listening

Scylla's Band
Benchmarks

Reproducible WER and listening-quality reports across voices, languages, and affect conditions. Every score stays connected to the generated audio and ASR diff.

Published releases

Shipping model v1

Ten voices, each voice's native English dialect, Spanish, Italian, and the complete v1 core affect matrix. The report shows coverage explicitly while a benchmark run is still being assembled.

Open v1 results →