TECHNOLOGY AND ATTRIBUTIONS
Olympia's spoken voice
Olympia is an AI guide. Her speech is synthesized by software, not a human narrator or a supernatural presence. The selected voice may sound different in different languages. These credits acknowledge the speech technology and relevant training data used for the site's offline neural voices.
English
Primary English voice: Kokoro, the af_heart synthetic voice. When unavailable, the site can use the local Piper en_GB-cori-medium voice.
Español / Spanish
Local voice: Piper es_ES-sharvard-medium, female speaker. The voice's documented dataset originates from the Edinburgh speech data collection and uses a Creative Commons Attribution 3.0 licensed dataset. See its model card and files.
Français / French
Local voice: Piper fr_FR-siwis-medium. The voice's documented dataset originates from Edinburgh DataShare and is licensed under Creative Commons Attribution 4.0. See its model card and files.
Voice recognition and accuracy
English microphone recognition uses a dedicated local English speech model. Spanish and French microphone recordings are processed privately within Olympia’s infrastructure using Whisper.cpp and the compact multilingual OpenAI Whisper model. The resulting transcript may contain errors. Visitors are asked to review and correct Spanish and French transcripts before Olympia answers. No system guarantees perfect recognition.
Conversation language selection influences both Olympia's answer and spoken voice. Translation quality, idioms, accents, and availability can still vary. Her foundational texts remain authoritative only in their published original form; AI-generated reflections are labeled accordingly.