SkillFoxx

VoxEasy skill

Skill for Codex and Gemini: turns a phrase, text or SRT into a Vox-style storyboard with prompts and JSON for Google Flow

  • 126 ★ on the repository
  • Last commit
Skill

Minimal access

Why this level

  • The skill only writes text prompts and JSON and runs nothing.
All reasons and checks

louchi1984-coder/voxeasyAuthor:louchi1984-coder

Install

In your terminal, with SkillFoxx CLI

npx skillfoxx add skills/voxeasy

Detects the agents on your machine, shows the access level and pins the version.

Other ways to install

Access and checks

How we review

Minimal accessThe tool runs nothing and makes no network calls.

  • The skill only writes text prompts and JSON and runs nothing.

This is third-party code. Review the repository files before installing.

What it does

VoxEasy takes a single phrase, a finished script or SRT subtitles and produces a storyboard for video generation in Google Flow: per-shot prompts and structured JSON. First the subtitle timeline is locked, each shot length snaps to the nearest 4, 6, 8 or 10 second step, and pauses between lines are preserved. For each shot the skill picks a way to show it: a direct picture of the subject, a story scene or a visual metaphor. Concrete objects, processes and data are shown directly and metaphor is kept for abstractions. Then the model, aspect ratio and style are confirmed. There are five public styles (`vox-standard`, `american-comic`, `monument-pastel`, `vintage-newspaper`, `pixel-theater`) and a `custom` mode. Hex colors steer the palette but never appear on screen.

Who it is for. Creators of short explainer videos who assemble clips in Google Flow and want ready per-shot prompts.

Good fit when

  • There is an SRT file and a shot-by-shot storyboard with 4, 6, 8 and 10 second lengths is needed.
  • The idea is one sentence and the script and shots must be expanded.
  • A Vox paper-cut collage video needs consistent styling for every shot.
  • News, product and data videos use separate skill variants.

Not a fit when

  • You need the video file itself: the skill outputs only prompts and JSON.
  • The video is assembled outside Google Flow.
  • Shots are not tied to subtitle timecodes.

Example request

Use $voxeasy: take this SRT, lock the timecodes and suggest a style for the explainer video.

Limitations

The skill does not create video: the result is text prompts and JSON, and you run generation in Google Flow yourself. The description and examples are in Chinese. Variants (`voxeasy-news`, `voxeasy-product`, `voxeasy-data`, `voxeasy-lab`) are built by `scripts/build_variants.py` from a shared core and installed by copying from `dist/`. The stable public version is connected by cloning the repository. The repository does not state a license. The Vox style belongs to Vox Media; the skill reproduces design techniques and not the publisher's material.

How to disable. Delete the `~/.codex/skills/voxeasy` or `~/.gemini/config/skills/voxeasy` folder. Saved storyboards stay in your working folder.

FAQ

How do I install VoxEasy in Codex?

Run `git clone https://github.com/louchi1984-coder/voxeasy.git ~/.codex/skills/voxeasy`. If the folder exists, back it up first.

Does it work with Gemini?

Yes. For Antigravity and Gemini clone the repository into `~/.gemini/config/skills/voxeasy`.

Why shot lengths of 4, 6, 8 and 10 seconds?

Google Flow generation yields clips of those lengths. The skill stores both the actual duration and the rounded-up step.

Which variants exist?

The stable `voxeasy` and the `voxeasy-news`, `voxeasy-product`, `voxeasy-data` and `voxeasy-lab` variants built from a shared core.

What is the custom mode?

The user describes a style in words and the skill builds a separate profile of composition, material, motion and camera for it.

Discussion

No questions or reviews yet.

Skills

A skill that turns one topic into a finished Vox-style paper-collage video: script, keyframes, motion, voice-over, music and captions

SkillExtended access2.8K

Skill turns finished SRT subtitles into a pack of image and video prompts: 21 styles, timecode storyboard, no media generation

SkillMinimal access125

Claude Code skill that turns one topic into a 60-second Vox-style explainer: script, keyframes, animation, voiceover and music.

SkillExtended access220

A skill that builds a narrated paper-collage explainer video from a topic, a talking-head clip or a single photo

SkillBroad access347

A Codex skill that renders images, SVGs and scripts as hand-drawn whiteboard videos, with 30 styles and narration

SkillExtended access350

A skill that turns your narration and real assets into a 20–60 second video in the style you pick, rendered locally as MP4.

SkillExtended access53

README in short

Skill output: prompts for Google Flow and JSON with actual and rounded shot durations. Variants differ in profiles while the core is shared. Documentation is in Chinese.

README badge

Original SKILL.md

---
name: voxeasy
description: 将一句话主题、完整文案或 SRT 制作为 Google Flow 可用的 VoxEasy 视频 Prompt 与 JSON。先规划字幕 Shot，再确认模型、比例、表达方式和所选风格。用于通用 Vox 风短视频，不用于静态海报、纯字幕校对、软件开发或视频剪辑。
---

# VoxEasy 4.13-public.1

严格执行：**先字幕 Shot，后参数选择；先选择表达方式，后风格转译；确认画面后才输出 Prompt。**

每次使用先完整读取 [`references/self-learning.md`](references/self-learning.md)，运行 `python3 scripts/self_learning.py validate` 检查本地学习状态。候选改进不是正式规则；只有通过学习门槛、备份、最小补丁和回归验证并记录到 `applied_patches` 后才可影响本流程。状态不可写时继续核心任务，并按输出合同返回最小 `Learning Observation`。

## 固定流程

1. 接收一句话、完整文案或 SRT；时效性事实与数据先检索验证。
2. 完整读取 [`references/timeline-rules.md`](references/timeline-rules.md)，按语义和连续时间区间规划 Shot，并向上吸附为 `4/6/8/10` 秒。
   SRT 计划完成后运行 `python3 scripts/validate_timeline.py <plan.json>`；修正全部错误后再确认。回归用例见 `evals/`。
3. 第一次确认字幕、Shot 边界、原始时间、空白和吸附时长；确认前不设计画面。
4. 让用户明确选择模型、实际生成比例和视觉风格。默认值只能推荐，不能视为已选。
5. 完整读取 [`references/expression-routing.md`](references/expression-routing.md)，逐 Shot 执行：字幕含义 → 核心信息 → 表达方式 → 场景锚点或隐喻映射 → 所选风格转译 → 构图、文字、动作、运镜。
6. 第二次确认完整视觉设计；确认前不输出最终英文 Prompt。
7. 完整读取 [`references/output-contract.md`](references/output-contract.md)，输出声画表、逐 Shot Prompt 与 JSON。用户确认成功或进入无关的新任务后运行 `python3 scripts/self_learning.py record-success`；累计 5 次成功使批量复盘到期，但不自动修改 Skill。

交互模式在两次确认和参数选择节点停止。只有输入明确带有批准状态与完整参数时，才可无头执行。

## 运行中反馈闭环

- 真实使用失败：在重试前运行 `record-failure`，立即执行一次 `review` dry-run。记录原因与可复用证据，但单次偶然失败不得直接生成或应用补丁。
- 用户明确纠正：先修正当前任务，同时运行 `record-correction`，立即生成待门槛审查的候选改进；候选不得覆盖本轮已确认内容。
- 五次成功复盘：`successful_uses_since_review` 达到 5 时，在当前交付结束后检查失败、纠正和候选项；没有足够证据时只完成复盘或拒绝候选，不为凑周期修改文件。
- 应用学习：先运行 `self_learning.py backup` 备份全部目标文件并生成 manifest，再实施最小补丁；完成原有验证、相关示例或人工检查后，才运行 `record-applied`。验证失败则回滚并运行 `reject`。
- 每次复盘只处理本 Skill 职责内的原子规则。不得把状态记录、候选项或维护过程混入正常视频 Prompt、字幕或画面设计。

## 输入与标题

- 一句话：生成连贯口播文案与候选标题。
- 完整文案：默认保留原文，只做语义分组与时长估算。
- SRT：原时间轴优先，不归零、不删除空白、不自动增加标题时长。
- 非 SRT 默认增加独立 4 秒标题 Shot，`voiceover_text` 为空；标题原样保留。
- 用户明确要求“测试 Prompt”或“只做一个 Shot”时进入快速测试：不自动补整条视频或标题；仍先展示该 Shot 的画面设计，除非用户明确要求跳过确认。

需要整条叙事结构时，按需读取 [`references/beat-architectures.md`](references/beat-architectures.md)；需要指定配色时，按需读取 [`references/palettes.md`](references/palettes.md)。

## 参数选择

- 完整读取 [`references/variant-profile.md`](references/variant-profile.md)；其默认值只用于推荐，用户明确选择优先。
- 推荐 `Google Flow / Omni Flash`、`9:16`、标准 Vox；不得自行扩展固定模型清单。
- 根据目标模型当前实际能力提供比例与时长，不得猜测。Flow/Omni 未确认原生支持 `1:1` 时，只提供 `9:16` 或 `16:9`。
- 用户需要方形成片时，使用其确认的实际比例生成，把全部关键内容放入中央 `1:1` 安全区，并在 JSON 记录 `crop_target_ratio: "1:1"`。
- 同一视频默认统一比例和风格；只有用户明确要求才逐 Shot 混合风格。

Report a problem