{
  "artifact": "demos/silero-aidar-xenia-dialogue-local.mp3",
  "origin": "local post-production from two local Silero v5_ru renders",
  "native_multi_speaker": false,
  "voices": [
    "aidar",
    "xenia"
  ],
  "source_audio": [
    "demos/silero-aidar-local.mp3",
    "demos/silero-xenia-local.mp3"
  ],
  "audio_duration_s": 9.36,
  "assembly_wall_s": 0.348,
  "sequence": "Aidar -> Xenia -> Aidar",
  "asr_check": "Здесь звук приходит короткими пакетами. Задержку ответа нельзя сводить к сети. Распознавание, сценарий и модель надо измерять отдельно.",
  "note": "This proves that one presentation artifact can contain two distinct local voices. It does not prove native multi-speaker inference by one model call."
}