Twitch & Kick VOD Analyzer: AI Scenes & Timestamps

Paste a public Twitch or Kick VOD or clip and get timestamped descriptions of what appears on screen. The analysis uses Marlin 2B (preset fast) through deAPI, so you do not need to bring a model key. Output is structured JSON in an Apify dataset, ready for highlight hunting, content research, or another AI agent.

deAPI video tools

• Need the spoken words? → Twitch & Kick VOD Transcriber: SRT, VTT & Speakers
• Analyzing TikTok, YouTube or X instead? → AI Video Analyzer: Scenes, Events & Timestamps

What does Twitch & Kick VOD Analyzer do?

It watches the picture of a stream recording and returns a time-ordered list of visual events: a layout change, a scoreboard, a facecam reaction, a menu screen, a new game scene, a title card. If the model also returns a scene summary, it is included. Each row carries the platform, the analyzed duration, the price charged, and the provider quote.

This is visual analysis only. The audio is not read, so no speech is transcribed.

What can AI detect in a stream?

• Scene and layout changes, such as a switch from gameplay to a lobby, a break screen, or a just-chatting overlay.
• On-screen moments with a timestamp: a match result, a menu, a shop screen, a cutaway, a product shot.
• People and objects that appear in frame, described in plain language.
• The order and timing of these events across the whole recording, so you can jump to the right minute.

Descriptions come from a model and can be imperfect. Check important timestamps against the original before publishing or acting on them.

Supported platforms and limits

• Twitch: VODs (twitch.tv/videos/<id>), clips (clips.twitch.tv/<slug>), and m.twitch.tv links.
• Kick: VODs (kick.com/<channel>/videos/<id>).
• Length: up to 2 hours (7,200 seconds) per video. Longer videos are rejected without any charge.
• Only public, accessible recordings work. Private, deleted, subscriber-only, or live streams can fail.
• Other sources (TikTok, YouTube, X) are not accepted here. Use the AI Video Analyzer instead.

Video analysis vs transcription

This Actor describes what you can see. A transcript captures what is said. They answer different questions, and many workflows use both: find the interesting visual moments here, then read the words around them in a transcript. For speech, use the Twitch & Kick VOD Transcriber and join the records by video URL.

How to analyze a VOD

1. Open the Actor and add one object per video to videos.
2. Set url to the public Twitch or Kick link.
3. Optionally set durationSeconds. If you leave it out, the Actor reads the length from the platform. If detection fails, that video returns a clear message asking you to provide it.
4. Optionally lower the two provider cost caps.
5. Run it and open the dataset.

{
  "videos": [
    {
      "url": "https://www.twitch.tv/videos/2868777996",
      "durationSeconds": 429
    }
  ],
  "maxDeapiCostPerVideoUsd": 1.0,
  "maxTotalDeapiSpendUsd": 2.0
}

Before each analysis the Actor obtains a provider price quote and skips any video that would exceed either cost cap. Long videos need a run timeout of roughly 1.3 times the video length plus a minute. The Actor skips a video without charge when the remaining run time is too short.

Output

One row per video, with fields such as:

• url, platform, status (success, skipped, or failed), and error when something went wrong.
• durationSeconds and events: the timestamped visual events, each with start and end times and a description.
• sceneSummary and scenes when the model returns them.
• apifyPricingEvent, apifyPricingEventCount, apifyEventPriceUsd, and providerQuoteUsd.
• analysisType: "visual_only" and audioTranscriptionIncluded: false, so downstream tools never mistake events for a transcript.

You can download the dataset as JSON, CSV, Excel, or HTML, or read it through the Apify API.

How much does it cost?

Pay per successful analysis. There is no Actor subscription and no model API key to buy. You are charged only after a result row is saved; failed or skipped videos have no analysis charge.

Item	Price (USD)
Per video	$0.0048
Per started minute of video	$0.0043
Actor start (once per run)	$0.00005

Examples:

Video length	Analysis price
7:09 (8 min)	$0.0392
36 min	$0.1596
1:48:52 (109 min)	$0.4735
2 hours (120 min)	$0.5208

The Actor is priced at cost (model provider cost plus Apify platform cost), with no markup. No minimum charge. The Store pricing tab is authoritative.

Try it on the Apify Free plan

Free-plan users get 3 analyses per calendar month, up to 20 minutes of material per video, and one video per run. The allowance resets at 00:00 UTC on the first of the month. Analysis and start events still use your Apify credit balance. Paying users have no monthly cap.

Use cases

• Find highlight candidates in a long VOD without scrubbing through it.
• Build a visual table of contents for a stream, with timestamps.
• Review a creator's overlays, scenes, and pacing across several recordings.
• Feed timestamped visual context to an AI agent or a clip-editing workflow.
• Pair with transcripts to get both what happened on screen and what was said.

FAQ

Can it analyze VODs longer than 2 hours? No. The limit is 2 hours per video. Split the VOD into parts, or analyze a clip of the section you care about.

Does it transcribe speech? No. It is visual analysis only. For speech, use the Twitch & Kick VOD Transcriber.

Do I have to provide the duration? No. It is optional for Twitch VODs, Twitch clips, and Kick VODs. If automatic detection fails, the Actor asks you to provide durationSeconds for that video.

Why was my video skipped? Common reasons are a provider quote above your cost cap, a run price limit that is too low, or too little run time left. The error field explains which one.

Can it analyze TikTok, YouTube, or X videos? Not this Actor. Use the AI Video Analyzer.

Is this affiliated with Twitch or Kick? No. This is an independent tool and is not affiliated with either platform.