Free neural narration

Kokoro-82M v1.0, by hexgrad. Model: Apache 2.0. ONNX conversion by ONNX Community, revision 4685882. Prepared recordings are produced offline using Kokoro CoreML by Jud (Apache 2.0), source revision 6bfc02a and models-2026-03-23. Older short samples used ONNX Runtime. The website only plays static audio files.

Kokoro licence · CoreML renderer licence · ONNX Runtime licence · Original voice catalogue

Italian: Sara / Nicola. English (US): Heart / Michael. Spanish: Dora / Alex. Portuguese (Brazil): Dora / Alex. French: Siwis (female only in this catalogue).

French voice dataset attribution: SIWIS French Speech Synthesis Database, University of Edinburgh, CC BY 4.0. Dataset and creators · Licence. The voice model is an adaptation; output is synthetic, not a recording of a live guide.

Text-to-phoneme preparation uses eSpeak NG through kokoro-onnx locally during publication. No third-party inference API is called, no API key is embedded, and no subscription is needed for this audio feature.

Tour content sources are linked beside each transcript. Italian scripts are original; translations were generated locally, with an editorial pass on English and targeted checks on Portuguese, Spanish and French. They have not had professional translation review. The narrated descriptions concern exterior visits; admission, opening hours and access are not included.