Find the moment a tutorial narrator says a phrase

What was saidTutorialVerified

0:58CC-BY-3.0 · Living Colour Filmproduktion, Hamburg, for Wikimedia Deutschland

Proof frames

Click a frame to seek

Tool calls

4 calls · agentic-video 0.3.1

transcribeimac · 62 ms

Arguments

{
  "source": "$CLIP"
}

Result

{
  "ok": true,
  "data": {
    "status": "in_progress",
    "requested_model_size": "tiny",
    "requested_model_cached": false,
    "transcription_status": "started",
    "retry_after_seconds": 15,
    "transcription_message": "The tiny transcription started in the background. About 0% done (estimate): 0 s elapsed of roughly 10 s. Retry in about 15 seconds.",
    "elapsed_sec": 0,
    "estimated_total_sec": 10,
    "estimated_percent": 0
  }
}

transcribeimac · 2 ms

Arguments

{
  "source": "$CLIP"
}

Result

{
  "ok": true,
  "data": {
    "status": "in_progress",
    "requested_model_size": "tiny",
    "requested_model_cached": false,
    "transcription_status": "in_progress",
    "retry_after_seconds": 15,
    "transcription_message": "A transcription for this video is already running. About 30% done (estimate): 3 s elapsed of roughly 10 s. Retry in about 15 seconds.",
    "elapsed_sec": 3,
    "estimated_total_sec": 10,
    "estimated_percent": 30
  }
}

transcribeimac · 2 ms

Arguments

{
  "source": "$CLIP"
}

Result

{
  "ok": true,
  "data": {
    "source_path": "~/hill/gallery/repo/cards/said-tutorial/clip.mp4",
    "language": "en",
    "duration_sec": 58,
    "segment_count": 15,
    "segments": [
      {
        "id": 1,
        "start_sec": 0,
        "end_sec": 2.56,
        "text": "free and cyclopedia written by its users.",
        "start_timecode": "00:00:00.000",
        "end_timecode": "00:00:02.560",
        "confidence": 0.908
      },
      {
        "id": 2,
        "start_sec": 2.56,
        "end_sec": 3.96,
        "text": "People like you?",
        "start_timecode": "00:00:02.560",
        "end_timecode": "00:00:03.960",
        "confidence": 0.908
      },
      {
        "id": 3,
        "start_sec": 3.96,
        "end_sec": 8.32,
        "text": "Today, out of more than 250 million Wikipedia visitors,",
        "start_timecode": "00:00:03.960",
        "end_timecode": "00:00:08.320",
        "confidence": 0.908
      },
      {
        "id": 4,
        "start_sec": 8.32,
        "end_sec": 11.04,
        "text": "over 10 million articles have been contributed",
        "start_timecode": "00:00:08.320",
        "end_timecode": "00:00:11.040",
        "confidence": 0.908
      },
      {
        "id": 5,
        "start_sec": 11.04,
        "end_sec": 15.76,
        "text": "by more than 500,000 editors in over 250 languages.",
        "start_timecode": "00:00:11.040",
        "end_timecode": "00:00:15.760",
        "confidence": 0.908
      },
      {
        "id": 6,
        "start_sec": 15.76,
        "end_sec": 17.44,
        "text": "It's really easy to share your knowledge",
        "start_timecode": "00:00:15.760",
        "end_timecode": "00:00:17.440",
        "confidence": 0.908
      },
      {
        "id": 7,
        "start_sec": 17.44,
        "end_sec": 20.16,
        "text": "with the many Wikipedia users around the world.",
        "start_timecode": "00:00:17.440",
        "end_timecode": "00:00:20.160",
        "confidence": 0.908
      },
      {
        "id": 8,
        "start_sec": 20.16,
        "end_sec": 23.88,
        "text": "Right now, you can help improve the world's biggest and best",
        "start_timecode": "00:00:20.160",
        "end_timecode": "00:00:23.880",
        "confidence": 0.908
      },
      {
        "id": 9,
        "start_sec": 23.88,
        "end_sec": 26.36,
        "text": "free and cyclopedia found on the net.",
        "start_timecode": "00:00:23.880",
        "end_timecode": "00:00:26.360",
        "confidence": 0.908
      },
      {
        "id": 10,
        "start_sec": 26.36,
        "end_sec": 29.68,
        "text": "Let me show you just how easy it is.",
        "start_timecode": "00:00:26.360",
        "end_timecode": "00:00:29.680",
        "confidence": 0.908
      },
      {
        "id": 11,
        "start_sec": 29.68,
        "end_sec": 34.32,
        "text": "When you search for something on the internet, one of the first hits you usually get is Wikipedia.",
        "start_timecode": "00:00:29.680",
        "end_timecode": "00:00:34.320",
        "confidence": 0.975
      },
      {
        "id": 12,
        "start_sec": 35.6,
        "end_sec": 40.16,
        "text": "So maybe you want to start your search directly in Wikipedia. Have a look at this.",
        "start_timecode": "00:00:35.600",
        "end_timecode": "00:00:40.160",
        "confidence": 0.975
      },
      {
        "id": 13,
        "start_sec": 41.36,
        "end_sec": 46.48,
        "text": "You'll find that most of the articles are comprehensive in up to date. However, you might come across",
        "start_timecode": "00:00:41.360",
        "end_timecode": "00:00:46.480",
        "confidence": 0.975
      },
      {
        "id": 14,
        "start_sec": 46.48,
        "end_sec": 51.44,
        "text": "an article that is too short or could use some improvement and who better to improve it than you.",
        "start_timecode": "00:00:46.480",
        "end_timecode": "00:00:51.440",
        "confidence": 0.975
      },
      {
        "id": 15,
        "start_sec": 52.16,
        "end_sec": 56.4,
        "text": "Wikipedia invites you to help build an even better encyclopedia.",
        "start_timecode": "00:00:52.160",
        "end_timecode": "00:00:56.400",
        "confidence": 0.975
      }
    ],
    "has_speech": true,
    "notice": "Note: Non-speech audio (music, sound effects, alarms) is NOT transcribed.",
    "confidence": 0.93,
    "uncertainty": "Whisper confidence is a no_speech_prob proxy; transcript content and alignment remain model-dependent."
  }
}

search_transcriptimac · 1 ms

Arguments

{
  "source": "$CLIP",
  "query": "right improve",
  "mode": "lexical",
  "top_k": 5
}

Result

{
  "ok": true,
  "data": {
    "hits": [
      {
        "segment_id": 8,
        "start_sec": 20.16,
        "end_sec": 23.88,
        "text": "Right now, you can help improve the world's biggest and best",
        "score": 0.8,
        "start_timecode": "00:00:20.160",
        "end_timecode": "00:00:23.880",
        "context": "Right now, you can help improve the world's biggest and best"
      }
    ],
    "total": 1
  }
}

Question and answer

Question: When does the speaker say "right improve"?
Answer: A keyword search for "right improve" first finds it 20.2 seconds into the clip.

Check passedanswer 20.16, expected 20.026 ±3.0

How it was checked

Truth from: captions
Fixed at: 2026-10-30 19:33

Made from the human captions published with the source (captions.vtt): the first caption line that starts at or after 20 s, ends inside the source, and has at least 6 words and 2 different words of 5 or more letters starts at 22.026 s of the source. The clip starts at 2.000 s of the source, so the line starts at 20.026 s in the clip; the truth is that time with a 3 s tolerance. No tool made this truth.

Keyword search only: the tool matched the words in its own speech-to-text, not their meaning. The clip is cut from a published video around one caption line, so the card checks that line only.

Try it yourself

Use agentic-video to find when my video says the words "right improve". Transcribe the video first, then search its transcript for those words in keyword mode (mode lexical), and tell me the start time of the first match in seconds. Use my own file.

Get the .mcpb file and open it in Claude Desktop.
Or: pip install agentic-video

Clip and licence

Licence: CC-BY-3.0
Author: Living Colour Filmproduktion, Hamburg, for Wikimedia Deutschland
Source: commons.wikimedia.org

This file is licensed under the Creative Commons Attribution 3.0 Unported license.