Audio semantic chapters

Before you run it4 local tools

Everything here is checked on your machine when rote inspects the play. This page never learns what you have installed or which credentials you hold.

• ○
  python3local tool

  Must be on your PATH. rote reports what it finds during inspect.

• ○
  ffmpeglocal tool

  Must be on your PATH. rote reports what it finds during inspect.

• ○
  ffprobelocal tool

  Must be on your PATH. rote reports what it finds during inspect.

• ○
  claudelocal tool

  Must be on your PATH. rote reports what it finds during inspect.

Inputs12, 1 required

rote asks for these before running and shows the defaults it will use.

• Audioaudio

  Path to the source audio or video file

  requiredstringcustom value allowed
• Max chaptersmax_chapters

  Maximum number of chapters to divide the recording into; fewer are used when the content has fewer natural divisions

  optionalstringdefault 16custom value allowed
• Out dirout_dir

  Directory for the MP3s; empty means a folder named after the source file, beside it

  optionalstringcustom value allowed
• Modelmodel

  Whisper model: tiny, base, small, medium, large-v3

  optionalstringdefault large-v3custom value allowed
• Languagelanguage

  Spoken language hint, or auto to detect

  optionalstringdefault autocustom value allowed
• Min gapmin_gap

  Minimum seconds between chapter starts; closer boundaries are merged

  optionalstringdefault 10custom value allowed
• Head secondshead_seconds

  Seconds re-transcribed at the head of each chapter during verification

  optionalstringdefault 8custom value allowed
• Work rootwork_root

  Scratch root for the inlined module, WAV mixdowns, transcript and analysis; cached per source file. Supports ~.

  optionalstringdefault ~/.cache/rote-audio-chapterscustom value allowed
• Whisper venvwhisper_venv

  Virtualenv holding the openai-whisper CLI (bin/whisper). Supports ~.

  optionalstringdefault ~/.local/share/rote-audio-chapters/venvcustom value allowed
• Claude binclaude_bin

  Claude Code CLI used for the semantic chapter division; needs its own authenticated session and spends that account's quota

  optionalstringdefault claudecustom value allowed
• Auto install whisperauto_install_whisper

  When true, create whisper_venv and pip install openai-whisper if absent instead of failing with instructions

  optionalstringdefault falsecustom value allowed
• Keep intermediatekeep_intermediate

  When true, keep the decoded WAV mixdowns in work_root. They are the bulk of the cache (~700MB per hour of audio) and take seconds to rebuild, while the transcript they sit next to takes minutes - so they are discarded after a successful run by default. Failed runs always keep them for debugging.

  optionalstringdefault falsecustom value allowed

What it changes