Yancey language courses · recording credits Synthesized dialogues and pronunciation models English, Spanish, French, German, Japanese, Korean, Arabic and Vietnamese use Supertone Supertonic 3, voice styles F2, M2 and F1, with F3 and M3 on the Vietnamese railway scene. Model license: OpenRAIL-M. https://huggingface.co/Supertone/supertonic-3 https://huggingface.co/Supertone/supertonic-3/blob/main/LICENSE French dialogue replies use hexgrad Kokoro v1.0, voice ff_siwis, rendered with kokoro-onnx. Model license: Apache 2.0. https://huggingface.co/hexgrad/Kokoro-82M https://github.com/thewh1teagle/kokoro-onnx Mandarin Chinese uses hexgrad Kokoro v1.1-zh, voice styles zf_001 and zm_010, rendered with sherpa-onnx. Model license: Apache 2.0. https://huggingface.co/hexgrad/Kokoro-82M-v1.1-zh https://huggingface.co/csukuangfj/kokoro-int8-multi-lang-v1_1 Earlier recordings Course dialogues and exercises: Yancey. Audio is synthesized. French: Piper fr_FR-mls-medium, trained on Multilingual LibriSpeech, OpenSLR 94 (CC BY 4.0), https://openslr.org/94/ German: Piper de_DE-thorsten-medium, Thorsten Voice (CC0), https://www.thorsten-voice.de/ Vietnamese: Piper vi_VN-vais1000-medium, VAIS-1000 (CC BY 4.0), https://ieee-dataport.org/documents/vais-1000-vietnamese-speech-synthesis-corpus Japanese: HTS Voice "Mei", MMDAgent Project Team, copyright 2009–2013 Nagoya Institute of Technology, Department of Computer Science. CC BY 3.0 https://creativecommons.org/licenses/by/3.0/ . Synthesized with Open JTalk. Mandarin Chinese (v2): Piper zh_CN-chaowen-medium, CC0 speech dataset, https://github.com/OHF-Voice/voice-datasets . Korean, Arabic and Mandarin Chinese (v1): synthesized with eSpeak NG, https://github.com/espeak-ng/espeak-ng Piper voice model cards: https://huggingface.co/rhasspy/piper-voices CC BY 4.0: https://creativecommons.org/licenses/by/4.0/ Phrasebook English: Piper en_US-libritts_r-medium, LibriTTS-R (CC BY 4.0). https://huggingface.co/rhasspy/piper-voices/blob/main/en/en_US/libritts_r/medium/MODEL_CARD Phrasebook Spanish: Piper es_ES-davefx-medium. https://huggingface.co/rhasspy/piper-voices/blob/main/es/es_ES/davefx/medium/MODEL_CARD Human-read passages in the Practice Studio Selected speech and transcripts from FLEURS (Few-shot Learning Evaluation of Universal Representations of Speech), Conneau et al., Google, 2022. Recordings are read by people. Original WAVs have been volume-normalized and encoded as mono MP3. The exercises, excerpt titles and answer choices were added by Yancey. Dataset and source recordings: https://huggingface.co/datasets/google/fleurs FLEURS license: Creative Commons Attribution 4.0 International, https://creativecommons.org/licenses/by/4.0/ Parallel texts originate from FLORES-101, Goyal et al., 2021, https://github.com/facebookresearch/flores The selected parallel texts and their adaptations are also available under CC BY-SA 4.0, https://creativecommons.org/licenses/by-sa/4.0/. FLEURS paper: https://arxiv.org/abs/2205.12446 Source-file index for each published recording: /reading-sources.tsv