{
  "input": {
    "text": "Why do AI agents' posts get rejected? Every social network measures length differently, and most agents count characters the wrong way.\n\nOn X the limit is 280, but it is a weighted count: each Chinese, Japanese or Korean character and each emoji counts 2, and every link counts 23 however long it is. A Chinese post is full at about 140 characters.\n\nBluesky allows 300 graphemes and caps a post at 3000 bytes. Farcaster counts UTF-8 bytes: 320 per cast, and a Chinese character takes 3 of them. Mastodon counts every link as 23 and defaults to 500. Threads by Meta allows 500 characters but counts an emoji as its UTF-8 bytes.\n\nThe fix: count the way each network counts, split at sentence boundaries, keep links whole, and check every post before publishing. Rules: https://github.com/twitter/twitter-text",
    "seconds": 30,
    "platform": "tiktok",
    "tone": "clear",
    "lang": "auto"
  },
  "output": {
    "platform": "tiktok",
    "network": "TikTok",
    "seconds": 30,
    "tone": "clear",
    "lang": "auto",
    "title": "Why AI agents' posts get rejected",
    "hook": {
      "text": "Why do AI agents' posts get rejected?",
      "seconds": 3.1
    },
    "totalSeconds": 26.7,
    "beatCount": 5,
    "beats": [
      {
        "n": 1,
        "start": 0,
        "end": 3.1,
        "seconds": 3.1,
        "voiceover": "Why do AI agents' posts get rejected?",
        "onScreen": "Why posts get rejected",
        "visual": "Close-up of a phone screen showing a social post with a red error under it"
      },
      {
        "n": 2,
        "start": 3.1,
        "end": 8.2,
        "seconds": 5.1,
        "voiceover": "Every network measures length differently, and most agents count the wrong way.",
        "onScreen": "Every network counts differently",
        "visual": "Split screen of several app compose boxes, each with its own character counter"
      },
      {
        "n": 3,
        "start": 8.2,
        "end": 15.7,
        "seconds": 7.5,
        "voiceover": "On X, each Chinese, Japanese or Korean character and each emoji counts 2, and every link counts 23.",
        "onScreen": "X: CJK + emoji = 2, link = 23",
        "visual": "A character counter jumping by two as emoji are typed"
      },
      {
        "n": 4,
        "start": 15.7,
        "end": 20.8,
        "seconds": 5.1,
        "voiceover": "Farcaster counts bytes: 320 per cast, and a Chinese character takes 3.",
        "onScreen": "Farcaster: 320 bytes",
        "visual": "Small blocks stacking up next to a short Chinese sentence"
      },
      {
        "n": 5,
        "start": 20.8,
        "end": 26.7,
        "seconds": 5.9,
        "voiceover": "The fix: count the way each network counts, and check every post before publishing.",
        "onScreen": "Count their way. Check first.",
        "visual": "Green check marks appearing beside a list of posts"
      }
    ],
    "caption": "Why AI agents' posts get rejected, and the fix: count length the way each network counts.",
    "hashtags": [
      "#AIagents",
      "#SocialMedia",
      "#ContentCreation"
    ],
    "srt": "1\n00:00:00,000 --> 00:00:02,800\nWhy do AI agents' posts get rejected?\n\n2\n00:00:03,100 --> 00:00:05,100\nEvery network measures length differently,\n\n3\n00:00:05,100 --> 00:00:07,900\nand most agents count the wrong way.\n\n4\n00:00:08,200 --> 00:00:10,600\nOn X, each Chinese, Japanese or\n\n5\n00:00:10,600 --> 00:00:12,600\nKorean character and each emoji\n\n6\n00:00:12,600 --> 00:00:15,400\ncounts 2, and every link counts 23.\n\n7\n00:00:15,700 --> 00:00:18,100\nFarcaster counts bytes: 320 per cast,\n\n8\n00:00:18,100 --> 00:00:20,500\nand a Chinese character takes 3.\n\n9\n00:00:20,800 --> 00:00:23,200\nThe fix: count the way each\n\n10\n00:00:23,200 --> 00:00:24,800\nnetwork counts, and check\n\n11\n00:00:24,800 --> 00:00:26,400\nevery post before publishing.\n",
    "pace": {
      "wordsPerSecond": 2.5,
      "hanPerSecond": 4,
      "kanaPerSecond": 6,
      "hangulPerSecond": 5
    },
    "warnings": [],
    "model": "example (hand-written to show the shape; paid calls are written by an open model on Cloudflare Workers AI)",
    "source": {
      "codePoints": 806,
      "example": true
    },
    "cached": false,
    "note": "Written by an open model on Cloudflare Workers AI from your source text, then timed and checked by Thread Kit. Durations are estimates from speaking rate (about 2.5 English words or 4 Chinese characters per second); on-screen text, title, caption and hashtags are kept short. Links not in the source were removed; numbers not in the source are listed in warnings. No video, voice or music is generated. Review before publishing."
  }
}