Saving a video is not the same as building a usable editing archive. A folder of numbered MP4 files tells an editor little about who made each clip, which download failed, or whether a file actually decodes. Video Link Downloader is useful at that handoff: it gives an agent a defined way to resolve authorized share links, save media and report what happened.

Open the Skill on OpenAgentSkill. This article covers the public Leon-Drq fork of Kunhai's original project, not a new video-generation model or an official Douyin integration. The listing is an owner publication; it is not an AI-review approval or a runtime certification.

What this Skill does—and what it does not

The pinned Skill instructions describe two routes. The Douyin archive script handles single links or a JSON batch, organizes files by creator, records outcomes, and checks downloaded media. The separate unified entry point also handles WeChat Channels, with browser dependencies.

The distinction matters for setup. Douyin archiving needs Python 3.10+, requests, FFmpeg and ffprobe. The WeChat Channels route additionally needs Playwright and Chrome. An installed SKILL.md is an instruction file, not a replacement for those programs.

Transcription is a separate, optional stage. The Skill describes extracting audio and using an available local speech-recognition system, but does not ship an ASR model or a transcription script. A parsed video title is not the spoken transcript. Nor does this tool create, edit or publish a new video on your behalf.

Example: brief an agent to build a creator archive

The following is an illustrative usage example, not a recording of an OpenAgentSkill test. No private media, live signed download addresses or fabricated success results are included. Supply a real public link only when you have the right to save its content.

Copy this task into your agent after installing the Skill and its dependencies:

Use video-link-downloader for the authorized Douyin links I provide. First confirm the destination folder and whether sending the share links to the documented third-party parser is acceptable. Save files by creator, preserve existing files, validate the media, and return the manifest and failed items. Do not transcribe, upload or repost anything unless I separately request it.

中文任务示例：

用 video-link-downloader 下载我提供且有权保存的抖音视频。先确认保存目录和第三方解析服务的使用，再按博主归档并校验音画。保留已有文件，返回视频清单、素材索引和失败项；不要自动上传、发布或转写。

The intended workflow is:

Authorized share links
        |
        v
Confirm source identity and creator
        |
        v
Resolve candidates -> Download -> Decode checks
        |                            |
        | failed                     | passed
        v                            v
Keep a failure record         Save creator archive
                                     |
                                     v
                         Optional local transcription
                         (separate setup and request)

For a batch already prepared as videos.json, run this from the reviewed repository directory:

python scripts/douyin_archive.py --input videos.json -o example

Each input object contains a url; author, title and a confirmed id are optional. Do not invent a creator when the link alone does not identify one. The script uses the uncategorized folder when no creator has been established. The repository usage guide includes installation and input guidance.

This is the expected output structure, not evidence that these files were generated during this article's preparation:

example/
  博主名/
    视频标题-ID.mp4
  视频清单.json
  素材索引.md

The archive is more useful than a bare MP4 because an editor can retain both the media and its processing record. Keep any permission records separately; a successful download is not proof of redistribution rights.

Choose the right operation before downloading

Your task	Suitable operation	What it does not establish
Inspect a share link	Use `--info`	Successful download or decoding
Save several authorized Douyin clips	Use `--input videos.json`	Permission to republish them
Improve editing compatibility	Consider `--h264`	Better source quality or universal editor compatibility
Extract spoken copy	Separate local ASR after download	A built-in transcription capability

The archive implementation checks container information and performs a full decode before delivery. It tries a bounded set of media candidates, rejects HTML masquerading as media, and avoids overwriting an existing target. A silent clip can be valid: missing audio is recorded rather than automatically treated as corruption.

On a repeated batch, the script can skip an archived item when its saved record, file size and container checks agree. That is not a universal duplicate detector: different links and missing IDs can still require inspection. Do not describe the archive as content-hash deduplication or assume that every retry performs another full decode of an already saved file.

Add transcription only when you need the spoken words

An editor may need a transcript for selecting quotes, writing subtitles or finding an opening hook. That is a good reason to add a second stage, not to silently expand every download request.

After a video has been validated, a separate workflow can extract mono 16 kHz WAV audio, run the chosen local Chinese ASR system, and save a matching _文案.md and _字幕.srt. Review names, English terms, numbers and negation against the audio. Keep the raw recognition output so corrections remain traceable.

Do not produce an empty transcript for a silent clip and call it successful transcription. Likewise, do not infer dialogue from the title or rewrite a speaker's claim as an established fact. If you only need a clip for editing, stop after the verified media handoff. For a later editing stage, see our guide to checking a trimmed video's first frame.

Limitations, privacy and operating cost

The Douyin flow sends the share URL to a third-party public parser. Its method notes identify sph.miuistore.com and a possible follow-up request to apikr.convry.com. Local storage therefore does not mean that the whole workflow is offline. Do not supply confidential or access-controlled links, and do not paste cookies or temporary signed media URLs into a public issue.

The documented Douyin path does not require an API key, browser login or a model call. That does not guarantee permanent free access to third-party services. Downloads consume bandwidth and disk space; decoding and optional transcoding consume CPU; local ASR needs its own model and compute. An agent host may also have its own subscription or usage charges.

Parser changes, connection failures or platform restrictions can interrupt the flow. Stop when authentication, payment or a verification challenge is required; do not try to bypass it. Keep failed items visible rather than declaring a partially completed batch successful. For important production work, retain an authorized original export as a fallback.

Selection methodology and source attribution

We selected this exact fork because its public Skill document matches the locally installed workflow requested by the site owner: Douyin HTTP parsing, creator-based batches and media validation. The reviewed snapshot is dda1cc61caf93b344552a7684e05def4bdccb84d. We inspected its instructions, README, implementation and method notes on October 19, 2026. We did not execute a live download or independently reproduce the author's historical download results for this article.

The repository retains Kunhai's MIT license and credits kunhai-88/video-link-downloader-skill as its upstream. The software license does not grant rights to videos obtained through it. This project is not presented as affiliated with Douyin, ByteDance, WeChat or the parsing services.

Start with the Skill's source and installation details, review the dependencies, and use one authorized clip before planning a batch. Judge the result by playable media and an honest processing record—not by whether an agent printed a download URL.