Video Scene Detection & Chapter Generator

Find every scene cut in a video and turn it into YouTube chapters, keyframes and a contact sheet. Paste links to your video files (MP4, MOV, WebM, MKV…, or Google Drive and Dropbox file links) and get, for every video:

• ✂️ Scene cuts: the start, end and length of every shot, frame-accurate, as JSON and CSV.
• 🖼️ Keyframes: one JPEG per scene, taken from inside the scene (never the first frame of a fade).
• 🗂️ Contact sheet: one image with up to 48 keyframes in a grid, each with its timecode, to see the whole video at a glance.
• 📑 YouTube chapters, ready to paste into a description (0:00 Introduction…), checked against YouTube's rules.
• 🧠 Optional AI chapters: the AI listens to the narration, starts each chapter where the topic changes (always at a real cut or pause) and writes a title and a one-sentence summary.

No API key, no software to install. Works for vlogs, tutorials, lectures, webinars, product demos, trailers and stock footage.

What can you do with it?

• YouTube chapters in seconds: paste the youtubeChapters text into your description; viewers jump straight to the part they want and Google can show the chapters as key moments.
• Video editing and logging: a shot list with timecodes and a picture of every shot, before you open the editor.
• Thumbnails and previews: pick a thumbnail from the keyframes, or use the contact sheet as a storyboard.
• Video libraries and AI pipelines: split long videos into scenes for search, tagging or feeding scene by scene to other tools.
• Quality checks: count shots, measure cut pace and find too-short scenes.

What videos does it accept?

Input	Example	Supported
Direct video file link	https://example.com/videos/demo.mp4	✅ mp4, mov, webm, mkv, avi, m4v… (H.264, HEVC, VP9, AV1…)
Cloud storage file URL	a public S3, Google Cloud Storage or CDN link	✅ if it opens without logging in
Google Drive file	https://drive.google.com/file/d/…/view	✅ shared as "Anyone with the link"
Dropbox file	https://www.dropbox.com/scl/fi/…?dl=0	✅
YouTube, TikTok, Vimeo or Instagram page links	https://www.youtube.com/watch?v=…	❌ link the video file itself

Up to 200 videos per run, each up to 240 minutes and up to 4K at 60 fps (set Max minutes per video to analyse only the start). A video that can't finish before the run's timeout is analysed from the start for as long as fits, and the record says so; raise the timeout for long videos. With the default 2 GB of memory a file can be up to about 3.9 GB; each extra 1 GB of memory adds 2 GB.

How much does it cost?

Pay per minute of video. No monthly fee.

Event	Price (FREE/BRONZE)	Lower on paid plans
Video minute analysed (cuts, keyframes, contact sheet, visual chapters, CSV)	$0.013 / started minute	down to $0.012
AI chapter minute (only with AI chapters)	$0.01 / started minute	down to $0.008
Actor start	$0.00005 / run	—

A 10-minute video costs $0.13, or $0.23 with AI chapters. Minutes are counted per video and rounded up, so a 61-second video is 2 minutes. Bigger or faster videos take longer to decode, so above 1080p at 30 fps a minute counts double (1080p at 60 fps, 4K at 30 fps) or four times (4K at 60 fps); larger videos aren't supported. Nothing is charged for a link that isn't a usable video, and if the AI can't make chapters for a video it gets visual chapters and the AI minutes aren't charged.

Cap your spend: set a maximum cost per run. When the budget can't pay for a whole video, only the minutes it covers are analysed, and the record says so.

Free plan: the AI runs on paid infrastructure, so free-plan runs include AI chapters for 10 minutes of video per run; the other videos get visual chapters. Scene detection, keyframes and visual chapters work on every plan.

How to make chapters for a video

1. Click Try for free. The example input analyses a 55-second sample video with 5 shots.
2. Paste your own video links into Video links, one per line.
3. Choose Chapters: Visual (at the strongest cuts, titled Chapter 1, 2, 3) or AI (titled from the narration). Set the Shortest chapter and Most chapters per video.
4. Run, then open the Videos view, the contact sheets, or download JSON, CSV or Excel.

Input example

{
  "videoUrls": ["https://example.com/videos/product-demo.mp4"],
  "chapters": "ai",
  "minChapterSeconds": 30,
  "maxChapters": 8,
  "chapterLanguage": "en"
}

Field	Default	What it does
videoUrls	—	Links to video files, one per line, up to 200.
chapters	visual	visual or ai.
minChapterSeconds	60	Chapters are at least this long (YouTube needs 10 or more).
maxChapters	12	The most chapters per video.
chapterLanguage	the video's own	Language of AI chapter titles and summaries.
sensitivity	3	How different a frame must be to count as a cut: lower finds more cuts, higher only hard cuts.
minSceneSeconds	0.6	Cuts closer together are merged, so flashes and fast montages don't become dozens of tiny scenes.
maxKeyframes	60	One JPEG per scene; with more scenes, they're picked evenly from start to end.
contactSheet	true	One JPEG with up to 48 keyframes and their timecodes.
maxMinutesPerVideo	120	Only the first this many minutes are analysed and charged.

Output

One record per video. Real output from the example input (shortened):

{
  "videoUrl": "https://api.apify.com/v2/key-value-stores/T5HtoeQYut1ufwkKC/records/short-cuts.mp4",
  "fileName": "short-cuts.mp4",
  "durationSeconds": 55,
  "analysedSeconds": 55,
  "width": 640,
  "height": 360,
  "fps": 25,
  "codec": "h264",
  "sceneCount": 5,
  "scenes": [
    {"index": 0, "start": 0, "end": 11, "duration": 11, "startTimecode": "0:00",
     "keyframeUrl": "https://api.apify.com/v2/key-value-stores/nV3tGrmsc3umTlfgs/records/short-cuts-64709a62aa-scene-0000.jpg"},
    {"index": 1, "start": 11, "end": 22, "duration": 11, "startTimecode": "0:11",
     "keyframeUrl": "https://api.apify.com/v2/key-value-stores/nV3tGrmsc3umTlfgs/records/short-cuts-64709a62aa-scene-0001.jpg"}
  ],
  "chapterMode": "visual",
  "chapters": [
    {"index": 0, "start": 0, "end": 11, "startTimecode": "0:00", "title": "Chapter 1", "summary": null},
    {"index": 1, "start": 11, "end": 22, "startTimecode": "0:11", "title": "Chapter 2", "summary": null}
  ],
  "youtubeChapters": "0:00 Chapter 1\n0:11 Chapter 2\n0:22 Chapter 3\n0:33 Chapter 4\n0:44 Chapter 5",
  "youtubeChaptersValid": true,
  "contactSheetUrl": "https://api.apify.com/v2/key-value-stores/nV3tGrmsc3umTlfgs/records/short-cuts-64709a62aa-contact-sheet.jpg",
  "scenesCsvUrl": "https://api.apify.com/v2/key-value-stores/nV3tGrmsc3umTlfgs/records/short-cuts-64709a62aa-scenes.csv",
  "billedMinutes": 1,
  "aiMinutes": 0,
  "warnings": [],
  "processingMs": 2328,
  "error": null
}

With AI chapters, a 2-minute narrated lecture got:

{
  "chapterMode": "ai",
  "chapters": [
    {"index": 0, "start": 0, "end": 30.6, "startTimecode": "0:00", "title": "What is a scene cut", "summary": "…"},
    {"index": 1, "start": 30.6, "end": 57.72, "startTimecode": "0:30", "title": "Why chapters matter", "summary": "…"},
    {"index": 2, "start": 57.72, "end": 81.64, "startTimecode": "0:57", "title": "The contact sheet", "summary": "…"},
    {"index": 3, "start": 81.64, "end": 107.04, "startTimecode": "1:21", "title": "Pricing and limits", "summary": "…"}
  ],
  "youtubeChapters": "0:00 What is a scene cut\n0:30 Why chapters matter\n0:57 The contact sheet\n1:21 Pricing and limits",
  "youtubeChaptersValid": true,
  "aiMinutes": 2
}

Times are seconds from the start of the video. A video that can't be used gets a record with only videoUrl and error (for example "video not available (HTTP 404): is the link public?") and isn't charged. Keyframes, contact sheets and CSV files are in the run's key-value store.

How accurate is it?

• Cuts: on a 1080p test video with a hard cut every 15 seconds, all 15 cuts were found at the exact frame, with no extra cuts. Videos larger than 1080p are analysed without their B-frames to halve the cost; there a cut can show up to 3 frames late.
• AI chapters: on a narrated lecture with 4 topics, every chapter started within 0.1 seconds of the real topic change, in each of 4 runs.
• Hard cuts are its strength. Slow fades and dissolves may be found a little late or not at all; lower the Cut sensitivity to catch more of them.

How fast is it?

Measured on Apify with the default 2 GB memory: a 4-minute 1080p video took about 85 seconds, a 2-minute 720p lecture with AI chapters 15 to 50 seconds (mostly waiting for the AI), and the 55-second example 2 seconds. Most of the time goes into decoding the video, so lower resolutions are faster. Videos in one run are processed one after another.

What it doesn't do

• Download from YouTube, TikTok, Vimeo, Instagram or other page links: link the video file.
• Cut the video into clips or export an edit list (EDL/XML) for an editor: use the scene times or the CSV.
• Recognise objects, faces or text in the frames.

Use it with AI agents (MCP)

https://mcp.apify.com?tools=bossless/video-scene-chapters

Ask your assistant to "make YouTube chapters for this video", and it runs this actor and reads back the chapters.

Use the API from code

curl -X POST "https://api.apify.com/v2/acts/bossless~video-scene-chapters/run-sync-get-dataset-items?token=YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"videoUrls": ["https://example.com/videos/demo.mp4"], "chapters": "visual"}'

from apify_client import ApifyClient

client = ApifyClient("YOUR_TOKEN")
run = client.actor("bossless/video-scene-chapters").call(run_input={
    "videoUrls": ["https://example.com/videos/demo.mp4"],
    "chapters": "ai",
})
for video in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(video["youtubeChapters"])

FAQ

Do I need an OpenAI or other API key? No. Scene detection and AI chapters are included in the price.

Why are my YouTube chapters marked as not valid? YouTube shows chapters only when the first starts at 0:00, there are at least 3, and each is 10 seconds or longer. A short video or one with few cuts may not have room for 3; lower Shortest chapter or use AI chapters.

Which languages work for AI chapters? Most major languages. Titles and summaries are written in the language spoken in the video, or in the one you choose.

What happens with a private or broken link? It's skipped with a clear message and not charged. So is a download that takes longer than 15 minutes. A run where every link was wrong still succeeds, with a message saying what to check.

Is my data stored or used for training? Videos are downloaded only for the run and deleted when it ends. Results are saved in your own Apify storage. For AI chapters, the video's audio goes to an AI model on Google Cloud (Vertex AI). Google's terms say it won't use this data to train or fine-tune its models; it may cache requests for up to 24 hours and log them for abuse monitoring.

Is this an official YouTube tool? No. This is an independent tool and is not affiliated with or endorsed by YouTube, Google or any video platform.

Credits

Scene detection uses PySceneDetect (BSD 3-Clause) and FFmpeg.

Related tools

• Speaker Diarization & Transcription: who spoke when for any audio or video file, with SRT and VTT subtitles.
• Call Recording Transcription & AI Call Analysis: transcripts, summaries and QA scorecards for call recordings.