Сквозь время · narration audio

The spoken lessons of «Сквозь время» (NoeticEcho/learn), served by GitHub Pages at https://open.noeticecho.space/learn-audio/, the same origin as the textbook.

The voice is synthetic. Every file here was rendered by Qwen3-TTS 0.6B CustomVoice (Apache-2.0), preset speaker serena, on the author's own machine. It is not a recording of a person, and it imitates no one.

What is here

• ru/<part>/<lesson>/<section>-<hash>.opus: one file per section of a lesson, Ogg Opus, mono, 32 kbps, loudness-normalised to −16 LUFS. <section> is intro (the title, the «Зачем» and the definition) or the section's number among the lesson's ## headings; <hash> is the hash of the words spoken in it, so a file never changes: a new text is a new file.
• en/<part>/<lesson>/<section>-<hash>.opus and en.json: the same for the English edition («Through Time»), read by the same preset speaker, serena, in English. Each language has its own tree and manifest.
• ru.json: the manifest, lesson id → sections → file, duration in seconds, the hash, and the start of each sentence. The textbook's build reads its own copy (content/narration/ru.json in NoeticEcho/learn) and offers «Слушать» only on a lesson whose every section matches its current text.

Nothing is built here and nothing runs: the files are rendered by tools/narration/render.py in NoeticEcho/learn and pushed. How, and how to re-render a lesson after its text changes, is in that repository's docs/narration.md.

Licence

The audio is a reading of the lessons and carries their licence: CC BY-SA 4.0, «Сквозь время», NoeticEcho. Voice model: Qwen3-TTS (Alibaba Qwen team), Apache-2.0.