商业产业 Hacker News Agent 2026-08-04T14:52 正文 ok 849 words

Show HN:OpenEdit——你的编程 Agent 现在可以编辑视频

英文标题:🔶 Show HN: OpenEdit – Your coding agent can now edit videos

OpenEdit 是一套开源、由 Agent 驱动的视频编辑流水线,可让 Claude Code、Codex 和 Gemini 等编程 Agent 完成剪辑、重构画面、添加字幕和动态图形等任务。它目前仅支持搭载 Apple Silicon 且运行 macOS Tahoe 26.0 的 Mac,Windows 和 Linux 版本仍在规划中。项目提供 VEED、WhisperX 本地离线转录及自定义转录服务三种方案,并要求提供逐词时间戳以避免字幕与音频不同步。其集成的 VEED HTML 渲染器据称在内部测试中最高比 Chrome 驱动的渲染器快 2.2 倍,但该结果尚未能在 VEED 外部复现。
打开原文相关 6 · 机会 1
media13.3 KB media12.7 KB media1.3 KB media1.4 KB media11.3 MB media66.5 KB media40.4 KB media59.3 KB media71.9 KB media71.4 KB media38.6 KB media53.5 KB media48.3 KB media47.0 KB media54.7 KB

中文

由 VEED 驱动的 OpenEdit

不是你租用的编辑器,而是你拥有的编辑器。

许可证:Apache-2.0 平台:Apple Silicon、macOS 26

安装 · 使用 · 示例 · VEED

OpenEdit 是一套开源、由 Agent 驱动的编辑流水线,内置 VEED 的 HTML 渲染器——该渲染器为闭源软件,但可免费使用。

它没有 GUI,也没有时间线。整个流水线完全由你的编程 Agent 驱动:编辑、剪切和重构画面;叠加动态图形及视觉元素;将幻灯片或网站转换为视频;捕获网页;以及在有帮助时接入任意视频/图像生成服务或 MCP 服务器。源视频并非必需:静态图片、幻灯片、生成式媒体或纯动态图形,在创作需求适合时都已足够。保持创意与灵活性——尝试新想法,而不是把每项需求都简化为字幕。

OpenEdit 发布视频

这支发布视频由 OpenEdit 制作——点击即可观看有声版本。

|
---|---
平台 | Apple Silicon Mac、macOS Tahoe 26.0
Intel Mac、较早版本 macOS | 不支持
Windows、Linux | 计划中;优先级取决于需求
转录 | 仅在需要为语音添加字幕时使用——OpenEdit 会询问一次并记住你的选择:VEED(注册 / 登录)、本地 WhisperX,或你自己的服务

下载此仓库,或通过命令行安装:

npx skills add veedstudio/open-edit

OpenEdit 使用与 Agent 无关的 skill 和仓库指南。它支持 Claude Code、Codex 和 Gemini;安装的 skill 会准备运行环境,并显式加载其 AGENTS.md 指令。

打开你的编程 Agent。例如,使用 Claude Code:

然后向它提出请求:

Add subtitles to my video [VIDEO]

Agent 将自动转录你的视频、分析转录文本、为视频创建合适的设计,然后进行渲染。

1 至 3 分钟后,Agent 会返回已烧录字幕的视频,并启动一个视频预览器。它还会告诉你最终 MP4 文件的位置。

随后你可以要求 Agent 修改字幕,例如:

I don't like the yellow colour, make it darker

Move the text up a bit

When he says "go buy it now", make sure the 'now' really stands out.

你也可以向 Agent 提供参考图片,让它模仿特定风格:

Add subtitles that look like this [IMG-REF], to my video [VIDEO]

以下为真实输出,每个输出均附有生成它的提示词。点击任一示例即可观看有声版本。

create viral subtitles with /open-edit and translate my video to 5 languages using VEED Lipsync 2.0 on Fal

一个源视频片段、三种语言、三种字幕风格:

generate 3 viral hooks in Seedance 2.0 on Fal and create dynamic motion graphics using /open-edit

三个 AI 生成的开场钩子、三种动态图形处理:

use Figma MCP to study my BrandBook and create branded campaign graphics using /open-edit

一份品牌手册、三张营销活动卡片:

字幕样式使用 HTML 和 CSS 编写。渲染不使用无头浏览器:无需在整个渲染期间安装、启动或维持浏览器运行。

内部基准测试显示,该渲染器最高比 Chrome 驱动的渲染器快 2.2 倍。这些测试结果仍属初步结果,尚无法在 VEED 外部复现;后续将提供有文档记录的基准测试及其方法论。

该渲染器已完成集成,但并非必需。合成流程不依赖于它,因此也可以改用 Chrome 作为渲染后端。

OpenEdit 会询问一次使用哪个服务提供商,并将答案记录在运行时根目录的 .open-edit-prefs.json 中;不会再次询问。每个服务提供商都会写入相同文件——runs/<key>/transcript.json,其中包含真实的逐词时间戳——后续流程无法分辨实际使用了哪个服务。

VEED ——质量最佳,也是默认工作流。需要 VEED 账户;使用限制遵循你的账户额度,免费账户每月约可转录 2 分钟,超出后需要购买套餐。为完成转录,音轨会被上传并存储;视频本身不会上传。默认情况下,OpenEdit 不会创建 VEED 项目。Agent 会引导你完成一次性浏览器登录,并将可刷新的令牌存储在运行时目录内的 veed/.veed-token.json。

node --import tsx veed/login.ts
node --import tsx veed/go.ts /path/to/video.mp4

WhisperX ——免费、本地、离线;任何内容都不会离开你的设备。按需安装,首次运行还会下载模型。在 Apple Silicon 上受 CPU 性能限制,并提供两个质量档位。

bash pipeline/scripts/install-whisperx.sh                  # on request, once
node --import tsx prep/transcribe.ts /path/to/video.mp4    # --model medium for the better tier

你自己的服务 ——可使用任意方式生成 Whisper 系列 JSON(WhisperX、openai-whisper、whisper-timestamped、mlx-whisper、带有 timestamp_granularities=["word"] 的 OpenAI API,或 whisper.cpp 的 -oj),然后进行映射。没有任何凭据会经由 OpenEdit 传递。

node --import tsx prep/whisper.ts transcription.json /path/to/video.mp4

无论使用哪一家服务提供商,都必须提供逐词时间戳:没有它们,字幕出现的时间会与音频逐渐失去同步。因此,没有逐词时间戳的转录文本会被拒绝,而不会以低质量方式渲染。

V1 的目标是字幕。动态图形、图表以及与品牌手册匹配的样式目前都可以渲染,但相较于字幕测试不足,预计会存在一些粗糙之处。

请通过 GitHub issues 报告缺陷。

该编辑器采用 Apache-2.0 许可证。渲染器二进制文件采用 PolyForm Shield 1.0.0 分发,该许可证允许你将制作的视频用于商业用途,且无需向 VEED 付款。完整条款请参阅 LICENSE 和 NOTICE。


OpenEdit · 由 VEED 驱动

English / 原文

OpenEdit, powered by VEED

Not the editor you rent, but the one you own.

License: Apache-2.0 Platform: Apple Silicon, macOS 26

Install · Usage · Examples · VEED

OpenEdit is an open-source, agent-driven editing pipeline that ships with VEED's HTML renderer — closed source, but free to use.

There is no GUI and no timeline. The pipeline is driven entirely through your coding agent — edit, cut, and reframe footage; layer motion graphics and visual elements; turn slides or websites into video; capture web pages; and pull in any video/image generating service or MCP server when it helps. Source video is optional: stills, slides, generated media, or pure motion graphics are enough when the brief calls for it. Be creative and fluid — try new ideas rather than collapsing every ask onto captions.

OpenEdit launch video

This launch video was made in OpenEdit — click it to watch with sound.

|
---|---
Platform | Apple Silicon Mac, macOS Tahoe 26.0
Intel Macs, earlier macOS | Not supported
Windows, Linux | Planned; prioritisation depends on demand
Transcription | needed only to caption speech — OpenEdit asks once and remembers: VEED (sign up / login), WhisperX locally, or your own service

Download this repo, or install via the command line with:

npx skills add veedstudio/open-edit

OpenEdit uses an agent-agnostic skill and repository guide. It supports Claude Code, Codex, and Gemini; the installed skill prepares the runtime and loads its AGENTS.md instructions explicitly.

Open your coding agent. For example, with Claude Code:

then ask it:

Add subtitles to my video [VIDEO]

The agent will automatically transcribe your video, analyse the transcript, create a suitable design for your video, and then render it.

After 1-3 minutes, the agent will return your video with subtitles burned in, as well as launch a video previewer. It will also tell you where to find the final MP4.

You can then ask the agent for subtitle amendments, for example:

I don't like the yellow colour, make it darker

Move the text up a bit

When he says "go buy it now", make sure the 'now' really stands out.

You can also hand the agent a reference image to copy a style from:

Add subtitles that look like this [IMG-REF], to my video [VIDEO]

Real outputs, each with the prompt that produced it. Click any example to watch it with sound.

create viral subtitles with /open-edit and translate my video to 5 languages using VEED Lipsync 2.0 on Fal

One source clip, three languages, three caption styles:

generate 3 viral hooks in Seedance 2.0 on Fal and create dynamic motion graphics using /open-edit

Three AI-generated hooks, three motion-graphic treatments:

use Figma MCP to study my BrandBook and create branded campaign graphics using /open-edit

One brand book, three campaign cards:

Caption styles are authored in HTML and CSS. Rendering does not use a headless browser: there is no browser to install, launch, or keep alive for the duration of a render.

Internal benchmarks measured the renderer up to 2.2x faster than Chrome-driven renderers. These measurements are preliminary and not yet reproducible outside VEED; a documented benchmark and its methodology will follow.

The renderer is integrated but not required. Composition does not depend on it, so Chrome can be used as the render backend instead.

OpenEdit asks once which provider to use and records the answer at the runtime root (.open-edit-prefs.json); it does not ask again. Every provider writes the same file — runs/<key>/transcript.json, with real per-word timings — and nothing downstream can tell which one ran.

VEED — best quality, and the default workflow. Requires a VEED account; usage limits are your account's, and a free account covers about 2 minutes of transcription a month, beyond which it needs a plan. The audio track is uploaded and stored in order to transcribe it; the video is not uploaded. By default OpenEdit does not create a VEED project. The agent guides the one-time browser login and stores a refreshable token at veed/.veed-token.json inside the runtime.

node --import tsx veed/login.ts
node --import tsx veed/go.ts /path/to/video.mp4

WhisperX — free, local and offline; nothing leaves your machine. Installed on request, and the first run also downloads a model. CPU-bound on Apple Silicon, with two quality tiers.

bash pipeline/scripts/install-whisperx.sh                  # on request, once
node --import tsx prep/transcribe.ts /path/to/video.mp4    # --model medium for the better tier

Your own service — produce a Whisper-family JSON however you like (WhisperX, openai-whisper, whisper-timestamped, mlx-whisper, the OpenAI API with timestamp_granularities=["word"], or whisper.cpp -oj) and map it. No credentials pass through OpenEdit.

node --import tsx prep/whisper.ts transcription.json /path/to/video.mp4

Per-word timings are required whichever provider you use: without them the caption reveals drift out of sync with the audio, so a transcript that has none is refused rather than rendered badly.

V1 targets captions. Motion graphics, charts, and brandbook-matched styling render today, but are less exercised than captions and should be expected to have rough edges.

Report defects through GitHub issues.

The editor is licensed under Apache-2.0. The renderer binaries are distributed under PolyForm Shield 1.0.0, which permits commercial use of the videos you produce with no payment to VEED. See LICENSE and NOTICE for the full terms.


OpenEdit · powered by VEED