voxflow 1.15.5 → 1.15.6

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "voxflow",
-  "version": "1.15.5",
+  "version": "1.15.6",
   "description": "AI audio content creation CLI — stories, podcasts, narration, dubbing, transcription, translation, and video translation with TTS",
   "bin": {
     "voxflow": "./dist/index.js"

package/skills/voxflow-slice/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: voxflow-slice
-description: "OFFLINE / LOCAL-ONLY slicing route — use when the user explicitly wants no-cloud / no-quota / offline / local slicing of a markdown / text article into a VoxFlow Slice deck.json. The user's own Claude does the slicing — there is no VoxFlow API call, no JWT, no quota deduction. Produces a validated 5–8 card JSON ready for `voxflow slice render deck.json` (local Remotion render). Pick this skill ONLY when the user says offline / local / no-cloud / 离线 / 本地 / 不联网 / no quota / 不用配额 — for cloud-route slicing (which calls VoxFlow backend and consumes quota), use the `slice` / `voxflow:slice` skill instead. Triggers: offline slice / local slice / 离线切片 / 本地切片 / no-cloud slice / 不联网切片 / slice this article offline / 文章离线转切片 / 本地切片视频."
+description: "Author + iterate on a VoxFlow Slice card video (1080×1920 vertical, 33 themes) locally in Claude Code / Cursor — the user's own Claude does the work, no VoxFlow API call to slice (0 quota). Full multi-turn workflow: (1) fork a curated template via `voxflow slice fork <id>` OR slice an article into a deck.json by hand; (2) edit deck.json fields in place with Edit/Write tools while `voxflow slice preview deck.json` is running (browser hot-reloads in ~50ms); (3) audition voice per card with ▶ in the browser (proxies /api/tts/synthesize, content-hash cached); (4) regenerate AI images per card with 🎨 (card.images: [{id, prompt, aspect, quality}], proxies hunyuan-image, content-hash cached); (5) render mp4 with `voxflow slice render deck.json` — audio + AI images baked in by default, --no-audio / --no-images opt out, --legacy silent fallback. Schema gains: per-card `voiceover` (voiceId/text/rate/enabled), per-card `images` registry, V2 LayoutTree `el: \"raw-html\"` escape hatch (4096 char cap). Loop rules: edit only the fields the user asked about (preserve diff highlight), never re-run `voxflow slice <article>` during iteration (overwrites edits), don't restart the preview server, don't call /api/audition or /api/imagine yourself (user-driven via ▶/🎨). Pick this skill for ANY iterative Slice work — even when the user has a token; the iteration loop is faster + cheaper than the one-shot cloud route. Triggers: slice / paper-slide / paperslide / card video / 卡片视频 / 切片视频 / 知识卡片 / 文章转视频 / 抖音知识号 / 小红书图文转视频 / 知乎长文转视频 / 公众号转视频 / iterate slice / multi-turn slice / 多轮编辑切片 / 迭代切片 / fork template / slice fork / 模板 / 模板市场 / template gallery / audition card / play voiceover / 试听口播 / 试听卡片 / regenerate image / 生图 / image-gen card / raw-html slice / custom panel / offline slice / local slice / 离线切片 / 本地切片 / 不联网 / no-cloud slice. For the one-shot cloud route (calls /api/slice/deck, 200 quota, no iteration), use the `slice` skill instead."
 ---
 
 # voxflow-slice — Article → deck.json (offline)