# Notices and acknowledgments Delibes-8M-esES modifications and token-ID runtime: Copyright 2026 Lapis Labs. Adapted from Paradee, Copyright Sahil Mahendrakar, Apache License 2.0: https://github.com/sahilmahendrakar/paradee Teacher: Kokoro-82M, hexgrad, Apache License 2.0: https://github.com/hexgrad/kokoro and https://huggingface.co/hexgrad/Kokoro-82M The student adapts the original English single-voice work to European Spanish, uses em_alex teacher targets and ships split deterministic-noise ONNX graphs. The original teacher weights, voice embedding and training states are not included. Kokoro's model card attributes upstream training audio to Koniwa, tnc (CC BY 3.0, https://github.com/koniwa/koniwa), and the SIWIS corpus (CC BY 4.0, https://datashare.ed.ac.uk/handle/10283/2353). Preserve these upstream acknowledgments: https://huggingface.co/hexgrad/Kokoro-82M https://creativecommons.org/licenses/by/3.0/ https://creativecommons.org/licenses/by/4.0/ These are upstream corpus credits, not proof of em_alex speaker consent. Training text also includes Tatoeba contributors, CC BY 2.0 FR; the per-sentence author/source/modification ledger is in third_party/tatoeba-attribution.csv. Imported material retains its original license. See third_party/TATOEBA.md. The project-authored sample texts and newly generated demo files are included with this Apache-2.0 review candidate, subject to the voice-rights caveat below. No GPL frontend, dictionary or engine is distributed. Separately installed NumPy, ONNX Runtime and SoundFile retain their own licenses; SoundFile's libsndfile dependency requires separate LGPL compliance when redistributing native binaries. Whisper base (MIT, OpenAI, https://github.com/openai/whisper) was used for evaluation only and is not needed or included at inference. The Apache license does not itself grant personality, publicity or voice rights. Voice-specific consent/provenance has not been independently established. This repository is a private review candidate, not commercial distribution clearance.