You know which ad won. Playhead tells you why.

POST /v1/watch_video
{
  "url": "https://www.facebook.com/ads/library/?id=…",
  "question": "Take this ad apart."
}

200
{
  "moments": [
    { "t": 0.0, "text": "A hand, two sachets, pink flowers" },
    { "t": 1.0, "text": "Text: AG1 Pro is HERE" },
    { "t": 2.8, "text": "Cut to a counter with the box" },
    { "t": 4.0, "text": "Text: Get your free Welcome Kit" }
  ],
  "session_id": "ses_…"
}

• Ads analyzer

  An ad in, the card out: the hook, the beats, the claims and the ask, each with its second. A set of ads becomes one grid.

  4 ads · 5,710 credits

  • 
  • 
  • 

• Watch video

  A link in, the answer out. One question reads the whole video and every line carries its second.

  1 question · 150 credits

  • 
  • 
  • 

• Video to prompt

  A prompt that remakes the video, shot by shot, with the second each shot starts.

  1 video · 1 prompt · 58 words

  • 
  • 
  • 

• SOP creator

  A screen recording in, a numbered guide out, with the button read off every step.

  1 recording · 3 steps

  • 
  • 
  • 

What people ask before they paste the first ad.

Can it tell me why an ad won?

Yes, from a set. One file on its own cannot: a video holds no spend, no ROAS and no retention curve. What it holds is the composition, and that is measured, down to the pace and the second of every claim. So mark the ads that won, add two that did not, and the answer is what the winners share and the others do not. That is the why, and it is read off the ads rather than guessed from your numbers. Your numbers are quoted beside it, never invented.

What does it read off an ad?

The hook, and how it reaches a viewer. The beats, each with its seconds. Every claim, quoted at the second it is legible. The offer, the ask, and whether the ad works with the sound off. Beside those, the numbers the engine counts rather than reads: how many cuts, how many a second, the median shot, and the second the product and the first face appear.

Can I compare a set?

Yes. Paste a column of links, up to ten. Every ad is read on its own, and the grid puts the same question to each of them. A yellow bar marks the rows where they do not agree.

Can an AI agent watch a video?

Not on its own. A language model cannot decode a video file, so it answers from the title, description or transcript instead. Playhead watches the video and answers in words, with the second of every moment it names. It is an API for your own software, and a connector for the agent you already use.

Does it work with my own files?

Yes. In any client with a filesystem, such as Claude Code, Cursor or VS Code, pass a path like ~/Movies/cut.mov. The file is uploaded once and reused for every later call, so your own edits, screen recordings and exports are analysable without publishing them anywhere.

Do I have to install anything?

No. Playhead reads the video on its own servers and sends back the answer. That is what makes it work in Claude on the web and on phones, where a local script cannot run at all.

Playhead is a video intelligence API for AI agents.

You paste a link, or you point it at your own file, and it watches the video and writes down what happens in it, which an assistant like Claude can actually read. Your own software calls it over HTTP. An agent reaches the same engine through a hosted connector, so nothing has to be installed. Signing in with Google is used only to create and access your Playhead account: we receive your name and email address, and nothing else. Your videos are yours. What happens to them is set out in the privacy policy.