AI Viral Clip Cutter — Long Video to Shorts

AI Viral Clip Cutter — Long Video to Shorts

Turn one long video into ranked vertical short clips with AI-scored moments and word-synced captions for TikTok, Reels, and Shorts.

What it does

The AI Viral Clip Cutter takes a single long-form video — a podcast, interview, livestream, or webinar — and returns a set of short clips ready to post. It runs four steps in one pass:

1. Transcribes the full video with AI. 2. Scores each segment for viral potential and ranks the strongest moments. 3. Cuts the top moments into standalone clips at 9:16 (vertical), 1:1 (square), or 16:9 (landscape). 4. Burns word-synced captions ("karaoke" style) directly onto the video, in your chosen caption preset.

You give it one video URL; you get back a ranked list of clips with their scores, timestamps, and download links. You control how many clips it returns (maxClips) and the minimum score it accepts (minScore, on a 0-100 scale), so you can keep only the moments worth posting.

It runs on Apify and is pay-per-use: you pay for the runs you make, with no monthly subscription.

You need an OpenAI key. The transcription and moment-scoring steps call OpenAI, so you supply your own openaiApiKey. Without it, a run returns a single labeled SAMPLE row instead of real clips. That sample row is a dataset row, and the $0.20 charge is per dataset row, so a keyless preview run costs $0.20 plus the start fee. Your OpenAI usage is billed by OpenAI, separately from the Apify per-clip fee.

Who it's for

Content creators, podcasters, and faceless/clip channels — and the agencies that serve them — who repurpose long videos into vertical Shorts, Reels, and TikToks. If you publish long episodes and want short clips out of them without editing each one by hand, or without committing to an OpusClip/Vizard/Klap subscription, this is built for that workflow.

Quick start

You can run this actor two ways: as a standard Apify actor (UI, API, schedule) or through the Apify MCP server so an AI assistant like Claude or Cursor can call it for you.

Run it as an actor

Open the actor in the Apify Console, paste a direct video URL into the input, add your OpenAI key, set maxClips and aspectRatio, and click Start. A minimal input that produces real clips:

{
  "videoUrl": "https://your-cdn.com/podcast-ep-42.mp4",
  "maxClips": 5,
  "aspectRatio": "9:16",
  "burnCaptions": true,
  "openaiApiKey": "sk-..."
}

videoUrl must be a direct video file (.mp4/.mov/.webm). To feed a YouTube link, pair it with a YouTube downloader actor first. When the run finishes, the clip records appear in the run's Dataset and the rendered MP4 files in the run's Key-value store.

Connect it to Claude or Cursor (MCP)

The actor is exposed through Apify's hosted MCP server over Streamable HTTP. The endpoint is fixed:

https://mcp.apify.com/?actors=dami_studio/ai-viral-clip-cutter

Authenticate with OAuth or an Authorization: Bearer <APIFY_TOKEN> header. A typical client config (Claude Desktop, Cursor, and other MCP clients use this shape):

{
  "mcpServers": {
    "ai-viral-clip-cutter": {
      "url": "https://mcp.apify.com/?actors=dami_studio/ai-viral-clip-cutter",
      "headers": {
        "Authorization": "Bearer YOUR_APIFY_TOKEN"
      }
    }
  }
}

Get your token from Apify Console → Settings → Integrations → API token. Once connected, you can ask your assistant something like *"cut my latest podcast at this URL into 5 vertical TikTok clips with captions"* and it calls the actor for you.

Input

Output

Each run produces a dataset where every item is one clip, plus the rendered MP4 in the run's key-value store. A clip record includes:

• rank — position in the ranking (1 = strongest moment).
• score — the viral-potential score (0-100) used to sort and filter clips.
• title — a short label for the moment.
• reason — why the model picked it.
• hashtags — suggested hashtags for the clip.
• start / end / durationSeconds — the source timestamps the clip was cut from.
• aspectRatio — the shape it was rendered at.
• clipKey / clipUrl — the key-value-store key and URL for the rendered clip with captions burned in.

Clips come back ordered by rank, so the highest-scoring moments are at the top. A clip that fails to render comes back with ok: false and an error so you can see what dropped — that row is still a dataset row and is billed like any other (see Pricing).

Example use cases

• Turn a podcast into 9:16 TikTok shorts
• Cut an interview into captioned clips
• Make 16:9 YouTube clips from a webinar

Pricing

$0.20 per dataset row, plus $0.0000125 each time a run starts. Flat rate — no volume tiers, no plan gates.

Billed per event, and these two are the whole list:

• $0.0000125 per actor start (per GB of run memory, minimum one unit — $0.00005 at this actor's default 4 GB).
• $0.20 per result (dataset item).

Read that second line literally: the charge is per row written to the dataset, not per usable clip. In a normal run one row is one finished clip, so five clips cost $1.00. But three other things also write a row and are billed the same $0.20:

• a keyless run (no openaiApiKey, or no videoUrl) writes one labelled SAMPLE row;
• a run where nothing scored above minScore writes one "no clips produced" row;
• a clip that fails to render writes one ok: false row.

So a preview run with no key costs $0.20, and a five-moment run where one clip fails to render still costs $1.00. Set maxClips to control the ceiling. There is no monthly subscription, no per-minute charge on the source video however long it is, and no separate fee for the transcription pass, the scoring pass, the 9:16 reframe or the burned-in captions — one row charge covers all of it.

Transcription and moment scoring run on your own OpenAI key, billed to you by OpenAI on top of the per-row fee and not marked up here.

FAQ

Does a high score mean the clip will go viral? No. The viral score is a ranking heuristic — it sorts moments by how clip-worthy they look, so you review the best candidates first. It is not a guarantee that any clip will go viral.

How is this different from OpusClip, Vizard, or Klap? Those are established, capable subscription tools. The main difference here is the model: this runs on Apify and is pay-per-use, so there's no monthly fee, and it slots directly into Apify workflows, schedules, the API, and MCP-connected AI assistants. It's positioned on pay-per-use and Apify-native automation, not as strictly better than those tools.

Which platforms are the clips for? TikTok, Instagram Reels, and YouTube Shorts for 9:16; 1:1 for square feeds; and 16:9 for landscape YouTube clips.

Do I need an OpenAI key? Yes. Transcription and moment scoring run on OpenAI, so you set openaiApiKey. Without a key the run returns a single labeled SAMPLE row instead of real clips, and because billing is per dataset row that preview costs $0.20 plus the start fee — it is a preview of the output shape, not a free trial. You can optionally point at an OpenAI-compatible endpoint with baseUrl and pick the detectModel. Your OpenAI usage is billed by OpenAI, on top of the Apify per-row fee.

Can an AI assistant run this for me automatically? Yes. Connect it through the Apify MCP server (see Quick start) and Claude, Cursor, or another MCP client can call the actor with a video URL on your behalf.