plugin.json

{
  "$schema": "https://agent-plugins.org/schemas/1.0.0/plugin.schema.json",
  "name": "transcript-kit",
  "version": "1.0.0",
  "description": "Turn a recording you own into a timestamped transcript with SRT and VTT captions, search what you have transcribed, find the best short clips in a talk, and translate the captions with a dub script.",
  "author": {
    "name": "Agent Tools",
    "email": "support@openkrill.app",
    "url": "https://tools.openkrill.app/"
  },
  "homepage": "https://tools.openkrill.app/",
  "extensions": {
    "com.openai": {
      "interface": {
        "displayName": "Transcript Kit",
        "shortDescription": "Transcribe audio and video",
        "longDescription": "Turn a recording you own into a timestamped transcript with SRT and VTT captions, search what you have transcribed, find the best short clips in a talk, and translate the captions with a dub script.\n\nAsk \"transcribe this podcast episode\", \"make subtitles for my video\", \"where did I mention pricing?\", \"find the best 60 seconds of this talk\" or \"translate these subtitles to Spanish\".\n\nTranscript Kit works on audio and video the user owns or has the rights to use: a file uploaded to the chat, or a direct link to an audio or video file. It does not download from YouTube, TikTok, Douyin or any other platform, and it does not read pages of those sites. Each file can be up to 15 MB and 15 minutes long (mp3, m4a, mp4, wav, ogg, flac or webm).\n\nSpeech is transcribed with the Whisper large-v3-turbo model on Cloudflare Workers AI, which costs about $0.0005 per audio minute. It is free to use, with no sign-up. To keep it free, each user and the service as a whole can transcribe only a limited number of minutes per day (UTC); every answer says how many minutes are left.\n\nEach transcript is kept for 24 hours so it can be searched and clipped, then deleted, and a library holds the latest 10. The media file itself is never kept. The search finds passages and repeated themes in your own transcripts only. Clip suggestions are text with start and end times and a hook line; no video is cut.\n\nCaptions can be translated into about 100 languages with the m2m100 model on Cloudflare Workers AI (about $0.0002 per 1,000 characters), keeping the original times, with a dub script of natural spoken lines sized to each segment. A call takes up to 18,000 characters, and translated text is not stored.\n\nA transcript is a model's best reading of the audio and can contain mistakes, especially with names, accents or noise. Check it before you publish it.",
        "developerName": "Agent Tools",
        "category": "Productivity",
        "capabilities": [
          "Transcribes an uploaded audio or video file or a direct file link, with timestamps",
          "Returns SRT and VTT caption files ready to use",
          "Accepts files up to 15 MB and 15 minutes in mp3, m4a, mp4, wav, ogg, flac or webm",
          "Searches your own transcripts for topics, hooks and repeated themes",
          "Suggests the best 30 to 60 second clips with start, end and a hook line, as text",
          "Translates captions into about 100 languages and writes a dub script sized to each segment",
          "Free with daily limits of minutes and characters; shows how much is left",
          "Keeps transcripts 24 hours, then deletes them; never keeps the media",
          "Does not download from YouTube, TikTok, Douyin or other platforms"
        ],
        "websiteURL": "https://tools.openkrill.app/",
        "privacyPolicyURL": "https://tools.openkrill.app/privacy",
        "termsOfServiceURL": "https://tools.openkrill.app/terms",
        "supportURL": "https://tools.openkrill.app/support",
        "defaultPrompt": [
          "Transcribe this recording and give me SRT captions",
          "Find the best 60 second clips in my talk",
          "Translate my captions to Spanish with a dub script"
        ],
        "brandColor": "#0F766E",
        "brandColorDark": "#14B8A6",
        "composerIcon": "./assets/icon.svg",
        "logo": "./assets/icon.svg"
      },
      "review": {
        "demo_recording_url": "https://tools.openkrill.app/demos/transcript-kit.mp4",
        "test_cases": {
          "positive": [
            {
              "description": "A transcript_id, the language, the length in seconds, the SRT text with numbered cues and clock times, the expiry time and the minutes left today. Check against: https://tools.openkrill.app/samples/clipper-talk.mp4",
              "prompt": "This is a talk our team recorded and published, so we have the rights to transcribe it. Transcribe it and give me SRT captions: https://tools.openkrill.app/samples/clipper-talk.mp4",
              "tools_triggered": "transcribe_media",
              "expected_behavior": "Asks the user to confirm they own the recording if that is unclear, then calls transcribe_media with url \"https://tools.openkrill.app/samples/clipper-talk.mp4\", confirm_rights true and include [\"srt\"]."
            },
            {
              "description": "Timestamped segments and captions, or a clear error that the link did not give an audio or video file or that the file is too large or too long. Check against: https://archive.org/download/short_poetry_001_librivox/because_I_could_not_stop_for_death_dickinson_64kb.mp3",
              "prompt": "This LibriVox recording is in the public domain, so I have the rights to transcribe it. Please transcribe it in English: https://archive.org/download/short_poetry_001_librivox/because_I_could_not_stop_for_death_dickinson_64kb.mp3",
              "tools_triggered": "transcribe_media",
              "expected_behavior": "Calls transcribe_media with url \"https://archive.org/download/short_poetry_001_librivox/because_I_could_not_stop_for_death_dickinson_64kb.mp3\", language \"en\" and confirm_rights true, because a public-domain recording may be transcribed by anyone."
            },
            {
              "description": "Passages with transcript_id, start and end times and a snippet around the word, or an empty result that says transcripts are deleted after 24 hours. Check against: The talk transcribed in the first case, which mentions backups at 0:34",
              "prompt": "Which of my transcripts mention backups, and at what time?",
              "tools_triggered": "search_transcripts",
              "expected_behavior": "Calls search_transcripts with query \"backups\" and the library_id from the earlier transcription."
            },
            {
              "description": "Spanish SRT and VTT with the original times, and a dub script with one spoken line per segment, each with start, end, max_chars and a fit of ok, tight or long. Check against: The talk transcribed in the first case",
              "prompt": "Translate the captions of that talk into Spanish and write me a dub script.",
              "tools_triggered": "translate_subtitles",
              "expected_behavior": "Calls translate_subtitles with the transcript_id of the talk and target_lang \"es\"."
            },
            {
              "description": "Up to two clips with start and end in seconds and as clock times, a hook line, the text and a rank score, and a note that no video is cut. Check against: The 51-second talk transcribed in the first case",
              "prompt": "Find me the best 2 clips of 10 to 20 seconds in that talk, with a hook line for each.",
              "tools_triggered": "suggest_clips",
              "expected_behavior": "Calls suggest_clips with the transcript_id of the talk, count 2, min_seconds 10 and max_seconds 20."
            }
          ],
          "negative": [
            {
              "description": "The plugin refuses platform pages and bulk downloads; it works on an uploaded file or a direct file link the user may use. Expected: Does not call any tool; says it only transcribes recordings the user owns or has rights to, one file at a time, and does not download from platforms.",
              "prompt": "Download every video from this TikTok creator and transcribe them all."
            },
            {
              "description": "A video page is not a media file, and the plugin does not fetch YouTube, TikTok, Douyin or similar pages. Expected: Does not call transcribe_media with the page link; explains that platform pages are not accepted and asks the user to upload their own file.",
              "prompt": "Transcribe this YouTube video: https://www.youtube.com/watch?v=abc123"
            },
            {
              "description": "The plugin returns text only: it cuts no video and posts nothing. Expected: Does not call any tool for editing or posting; says it only suggests clip times and text and can return captions, not edit or publish video.",
              "prompt": "Cut my video into a 60 second clip with subtitles burned in and post it to my channel."
            }
          ]
        }
      },
      "publication": {
        "release_notes": "First release. Transcribes uploaded files and direct file links (up to 15 MB and 15 minutes) with Whisper on Cloudflare Workers AI, returns segments, SRT and VTT, searches the user's own transcripts for passages and repeated themes, suggests 30 to 60 second clips as text, and translates captions with a dub script (m2m100 on Workers AI). Transcripts are deleted after 24 hours."
      },
      "apps": "./.app.json"
    }
  }
}