{# Settings → Synthetic speech detection (#145): the plugin's tri-state toggles (enabled + autogenerate). Follows the _semantic.html / _translation.html pattern: self-contained section with own form, rendered from _settings_context. The synthdetect service URL is env-only (no UI override). #}

Synthetic speech detection

Score each speaker turn for synthetic speech risk using a dedicated model service. Scores are calibrated risk values, not portable probabilities. The service must be running and reachable at the configured URL for scoring to succeed. Changes here apply to your next run or job, no restart needed.

{% for error in synthdetect_errors %} {% endfor %}
Synthetic speech detection

Score speaker turns for synthetic speech risk. Requires the synthdetect model service to be running.

Score new runs automatically

Score each run as it finishes. With this off, score individual runs on demand from the run detail page. Requires synthetic speech detection to be on.