@gitruck/cli 1.3.0 → 1.3.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/AGENT.md +17 -3
- package/README.en.md +2 -0
- package/README.md +2 -0
- package/dist/index.js +248 -248
- package/docs/reference.en.md +34 -0
- package/docs/reference.md +34 -0
- package/package.json +1 -1
package/AGENT.md
CHANGED
|
@@ -271,9 +271,23 @@ gtrk split <拆分稿.json> --project <目录> --md --json # ⑤ 校验落地
|
|
|
271
271
|
- **只动 `struct_meta.split`**:写回不碰 materials/tracks,配合客户端「保存 → 发起 → 写回 → 重载」闭环(opencut 联动)。
|
|
272
272
|
- **下游消费**:`dispatch.json` 的 `film_broll` 队列 → `gtrk matrix --project <dir>`(B-roll 双口检索 → `split/broll-plan.json` 候选清单;url 24h 过期,重跑即重签);`mg` 槽位表(去品牌化前 `rrv_mg`)→ `gtrk mg` 铺轨 / real-roam-viz 产颗粒;`ai_drama` 队列 → ai-drama-prompter。
|
|
273
273
|
|
|
274
|
-
> **skill 分工**:拆 beat / 选 lane / 写 handoff 是脑(`gtrk-splitter` skill)的活;投影 / 校验 / 落地 / 写时码是手(本命令)的活。skill 铁律:只引用视图存在的 utterance id、不抄原文定位、不碰时码。
|
|
275
|
-
|
|
276
|
-
### 2.
|
|
274
|
+
> **skill 分工**:拆 beat / 选 lane / 写 handoff 是脑(`gtrk-splitter` skill)的活;投影 / 校验 / 落地 / 写时码是手(本命令)的活。skill 铁律:只引用视图存在的 utterance id、不抄原文定位、不碰时码。
|
|
275
|
+
|
|
276
|
+
### 2.2 精修工程导出:`gtrk project export --project <目录>`
|
|
277
|
+
|
|
278
|
+
口播起盘用 `gtrk project init`,精修收尾用 `gtrk project export`;已经在客户端铺过 B-roll、MG 或 BGM 的 `.gtrk`,直接导出 NLE 工程,**不要重跑 `oralcut`**,否则会丢掉精修并重复计费。
|
|
279
|
+
|
|
280
|
+
```bash
|
|
281
|
+
gtrk project export --project ./my-video --formats jianying,xml
|
|
282
|
+
gtrk project export --gtrk ./my-video/gtrk/project.gtrk --formats xml --no-open
|
|
283
|
+
```
|
|
284
|
+
|
|
285
|
+
- 颗粒会随 NLE 工程一起导出:命令会预渲 qtrle;首次导出按分钟计费,客户端已经烤过或第二次导出命中 `.tonghe-cache/particles/` 时复用缓存。
|
|
286
|
+
- 本命令有两个计费面:工程结构按次,颗粒按分钟;这和 `project init` 的 0 积分不是一回事。执行前会分别提示实时价格。
|
|
287
|
+
- 只选择 `gtrk` 时会自动跳过颗粒预渲(gtrk 原生保留 `beat_track`),不要再手动加 `--no-particles`;`--no-particles` 只在明确接受 NLE 导出缺少颗粒时使用。
|
|
288
|
+
- `--json` 只在 stdout 输出一行结果 JSON;产物目录恒写 `result.json`,源 `.gtrk` 只读不改。
|
|
289
|
+
|
|
290
|
+
### 2.3 视频转文字稿:`gtrk transcript <本地视频>`
|
|
277
291
|
|
|
278
292
|
这一能力由独立的 `gtrk-transcript` Skill 驱动,不属于 `gtrk-tools` 单点工具族。用户说「视频转文字 / 视频转文字稿 / 提取视频文稿 / 把本地视频整理成妙记式文稿」时,直接运行:
|
|
279
293
|
|
package/README.en.md
CHANGED
|
@@ -159,6 +159,8 @@ flowchart LR
|
|
|
159
159
|
|
|
160
160
|
The three-format rough-cut output and the export of a fully packaged project have different support boundaries. The [workflow guide](docs/workflow.en.md) explains where each deliverable opens and which effects need rendering.
|
|
161
161
|
|
|
162
|
+
After refining a `.gtrk` project in the client, use `gtrk project export --project <dir>` to produce Jianying, Premiere XML, FCPXML or OTIO while preserving the refined timeline; do not rerun `oralcut`, which rebuilds the rough cut. Export renders MG particles into the NLE package: structural export is billed per task, and particles missing from the shared cache are billed by the minute. `project init` remains zero-credit. When only `gtrk` is requested, the native `beat_track` is kept and particle rendering is skipped.
|
|
163
|
+
|
|
162
164
|
## gtrk + AI Drama Desk: from storyboard to editable clips
|
|
163
165
|
|
|
164
166
|
For AI-generated story scenes, connect [Gitruck AI Drama Desk](https://github.com/Gitruck/ai-drama-desk) to the same project workflow. The responsibilities stay separate:
|
package/README.md
CHANGED
|
@@ -159,6 +159,8 @@ flowchart LR
|
|
|
159
159
|
|
|
160
160
|
粗剪阶段的三端工程,与完整包装后的导出支持范围不同;[工作流指南](docs/workflow.md)说明每种交付物在哪打开、哪些效果需要渲染。
|
|
161
161
|
|
|
162
|
+
客户端精修 `.gtrk` 后,用 `gtrk project export --project <目录>` 导出剪映、Premiere XML、FCPXML 或 OTIO,保留精修时间线;不要重跑 `oralcut`,否则会重新生成粗剪。导出会把 MG 颗粒预渲进 NLE 工程,结构导出按次计费,未命中共享缓存的颗粒按分钟计费;`project init` 仍是 0 积分。只选 `gtrk` 时会保留原生 `beat_track` 并自动跳过颗粒预渲。
|
|
163
|
+
|
|
162
164
|
## gtrk + AI Drama Desk:从分镜到可回轨片段
|
|
163
165
|
|
|
164
166
|
如果你的素材不是实拍,而是希望把一份分镜稿变成 AI 情景片段,可以把 [Gitruck AI Drama Desk](https://github.com/Gitruck/ai-drama-desk) 接到同一条工程链里。两者分工很清楚:
|