textcastlisted
Install: claude install-skill veardk/textcast
# Textcast
Pipeline: **Ingest → IR → Enrich → Render → Open**.
| Stage | Who | Output |
|---|---|---|
| Ingest | `scripts/textcast.py ingest` | `ir.json`: cues + sentence segments with real timestamps |
| Enrich | **you, directly, in this conversation** | `enrich.json` (languages, topics, analysis, summary) + `translations/*.json` |
| Audit | `scripts/textcast.py audit` | completeness / ordering / translation coverage check |
| Publish | `scripts/textcast.py publish` | `textcast/output/<category>/NN-<slug>.html` (mode 1) and/or `NN-<slug>-summary.html` (mode 2), plus the rebuilt `textcast/output/index.html` |
| Open | `npx -y textcast-chat-axi@latest` | the published page in a local reading window with the conversation panel and video player — always run after publishing |
Scripts do everything deterministic (fetching, parsing, segmenting, timing, coverage). You do
the language work **yourself, right here**: choosing languages, topic boundaries, paragraphs,
analysis, translation, summary. You are the translator — nothing else in this pipeline translates.
The data contract is in `references/ir-schema.md` — read it before writing `enrich.json`.
Run scripts with `python3 <skill-dir>/scripts/textcast.py …`. Every path below (`textcast/output/…`)
is relative to the folder you are working in, never the skill folder: the library goes to
`./textcast/output/` there, and the scripts create it on first use. Requirements: Python ≥ 3.10 and
Node ≥ 22 (the channel fetch scripts and `te