recite 发布 v0.1.0

这个版本更新了什么#

First numbered release of recite: text to verified speech through the local Codex ChatGPT sign-in. Every chunk is rendered by the realtime model, transcribed independently and checked against the text; a take that does not match is retried, and the run tells you what stayed unverified.

Install#

brew install xiaolai/tap/recite

Or download the archive for your platform below, check it against SHA256SUMS, and put recite on your PATH. The macOS binaries are signed with a Developer ID and notarised by Apple; the first run needs a network connection so Gatekeeper can fetch the notarisation ticket. recite needs the Codex CLI signed in with a ChatGPT account, and ffmpeg for encoded formats other than WAV and PCM (on macOS, m4a can also use afconvert).

Important#

recite relies on an undocumented use of the ChatGPT sign-in, which may stop working at any time. Usage is your responsibility under OpenAI's terms. See the README.

What is in 0.1.0#

say for plain text, Markdown with front matter, spoken scripts and multi-speaker dialogue scripts, to WAV, MP3, M4A/M4B with chapters, Opus, Ogg or FLAC, with subtitles and loudness presets
render for projects with many outputs
check and prepare to see what would be read, with no network
doctor, auth, usage, cache, config, voices, version (also v and --version) release-check-v0.1.0.md is the live release check this build passed: 10 of 10 steps.

以上是作者写在 GitHub 上的英文更新说明原文，未经翻译。查看原页面