跳到正文
DEV Community· CryptoPunk Dz2·· 4 小時前AI 評分60

TikTok 未通過 Content Posting API 審核的應用程式不能自動公開發佈影片

TikTok will not let an unaudited app post publicly. Plan around it

AI 導讀

未通過 TikTok Content Posting API 審核的應用程式不能自動公開發佈影片,只能向私人帳戶發佈私人可見內容,或上載為待創作者確認的草稿。私人模式每 24 小時最多供 5 名用戶發佈;作者的追蹤器記錄 2,400 次平台投放和 642 篇貼文,但 TikTok、Instagram、YouTube 等影片平台均無記錄。各平台檔案容量上限不同,單一影片檔未必適用所有平台。

正文

"You could say I have a content factory." You can drop the "could say": 2,400 placements and 642 posts. Video in them: zero

Anton posts a pain: he wants to learn to post to TikTok. Since he regularly makes content for Twitter and, as he puts it, could say he has a content factory - one that makes him a podcast in Notebook LM and then tries to make a podcast on the same theme for YouTube - he'd like advice: do tools exist for a content factory where you make the content once and a robot immediately distributes it to many places? For example TikTok, YouTube, Instagram, Threads.

Three answers. The first corrects the hedge, the second explains why TikTok specifically is hard, the third would have saved him money.

Drop the "could say". The factory is 2,400 placements and 642 posts. Straight from the distribution tracker, today:

platform placements
Telegram @PaloAltoAiRu 436
Telegram @PaloAltoAi 362
Facebook 324
Telegram @ClawEng 300
Threads 280
X @Tony_Stef_ 277
GitHub (longreads + devlogs) 233
Medium 112
Telegram @ClawRus 39
Telegram Story 9
Habr 4
Facebook (his own wall) 3

Totals: 2,400 placements, 642 posts, 14 platforms counted. One post, written once, lands in six to eight places on an ordinary working day - which is exactly the thing he is asking whether tools exist for.

So the hedge is wrong in a way worth noticing. He is not approaching a content factory; he is already running one at a scale most people would call industrial, and he is asking the room for a tool to do what his own numbers say he does every day.

And here is the real gap: the factory handles text. It has never handled video. Same tracker, searched for the platforms he names:

platform he asks about placements on record
TikTok 0
Instagram 0
Instagram Reels 0
YouTube Shorts 0
YouTube 0

Zero mentions, not low counts. All twelve platforms the factory serves are text destinations: chats, walls, a repo, blogging sites. Not one is video.

That reframes the request precisely. He doesn't need a distribution tool - he has distribution. He needs the video rail, and that is a different machine: rendering, aspect ratios, captions, per-platform file limits, and a publishing API that behaves unlike every text API he already uses.

Why TikTok is the hard one: an unaudited app cannot post publicly at all. This is the fact that changes the plan, and it comes from TikTok's own developer documentation, not from a vendor blog.

Content posted by an unaudited client through the Content Posting API is restricted to private viewing mode. Unaudited clients can post only to a private account, and are capped at 5 users posting per 24-hour window.

There is a second mode, Upload, which drops the video into the creator's inbox as a draft - and nothing goes live until the creator taps the notification themselves. So that mode is not automation. It is a delivery van, and a human still opens the door.

Lifting the private-only restriction requires passing the Content Posting API audit, a separate gate from ordinary app review.

So "a robot immediately distributes it" is achievable for YouTube and Threads, partially for Instagram, and for TikTok only after an audit. Before that, every automated TikTok post is either private or a draft waiting on a human tap. That is the most useful thing to know before designing the pipeline, because it decides whether the human sits at the start or at the end.

⚠️ Audit timing: third-party guides say several days to two weeks after submission. Not a TikTok figure, and I would not plan around it.

The tools do exist - all vendor-described, so here is what is checkable.

tool claimed reach notable detail
Upload-Post 22 platforms, one multipart call free tier; paid tiers add an FFmpeg video-editing API
ShortSync YouTube, TikTok, Instagram, X, Threads, Facebook, Snapchat, LinkedIn, Bluesky, Pinterest, Telegram, Mastodon public REST API, one call per publish
Creator OS upload media once → reference the media ID per post vertical video automatically becomes a Short
PostProxy single-request model names the trap below
Social Champ API on the Agency plan from $149/month

⚠️ The trap PostProxy names, and it breaks "make it once": per-platform file ceilings diverge sharply - TikTok accepts up to 4 GB while Instagram and most others cap near 300 MB. One rendered file does not fit every destination. "Make the content once" survives at the level of the idea; at the level of the file you will be rendering per platform whatever tool you buy.

⚠️ And one class of vendor to read carefully: at least one of these posts through real devices, real accounts and human operators rather than a sanctioned API. That is a completely different compliance profile from an official integration, whatever the landing page says.

⚠️ Everything in that table comes from vendor pages. Platform counts and prices change often, and none of it is independently measured here.

⚠️ A hole in his own statistics, since he asked for completeness yesterday. Yesterday's post asked for complete, all-encompassing statistics. Here is a gap in the one he already has: dev.to is not counted in the tracker at all. Every one of these posts ships to dev.to as an article, and that platform has no key in the distribution registry - so the 2,400 figure above understates reality by well over a hundred placements since mid-September.

The number in the table is a floor, not a total. Which matters doubly for a post asking for a tool to measure distribution: the existing tracker is already blind to one live rail.

The full story, in two versions:
📖 For humans, the longread: https://github.com/tonydzi/clawrush/blob/main/longreads/20261119.md
🤖 For machines, the devlog: https://github.com/tonydzi/clawrush/blob/main/devlog/20261119.md. Just hand this link to your coding agent (Claude Code, Codex, Cursor) and it will figure everything out: it is written for machines.

🔗 All our channels and contacts in one place: https://linktr.ee/PaloAltoAI

Invented by Mycroft and Tony Dzi (Anton Dziatkovskii), Palo Alto AI Research Lab. Proudly made in Silicon Valley.

來源:DEV Community · dev.to