VizPT takes a short video (up to 20 seconds) and returns an annotated version with numbered boxes on every moving subject. Each box follows its subject across the clip and is labeled with what the subject is and what it is doing. You can download the annotated video or grab any single frame as a PNG.

Yes. VizPT is free to use and does not require a signup. A rate limit applies per IP to keep the service running on a single server.

Can I record a video directly in VizPT?

Yes. Click the Record button to open the in-browser recorder. Your camera feed is captured locally via MediaRecorder, and you can preview and retake before analyzing.

What AI model powers VizPT?

VizPT runs a vision pipeline tuned for action extraction from consecutive frame pairs. The internal model is managed by us and subject to change without notice.

What video formats does VizPT accept?

VizPT accepts MP4, MOV, and WebM. The maximum size is 50 MB and the maximum duration is 20 seconds. Longer clips are truncated during sampling.

VizPT was built and is operated by Sandeep Kukreja. It is one of six live web products on his independent AI portfolio at sandeepkukreja.net.

VizPT — AI Video Action Extraction | Extract Verbs, Objects & Timestamps From Any Clip

Q: Does VizPT upload my video?

No. Your raw video file never leaves your browser. Only small, downgraded image samples go to our backend for analysis.

What VizPT does

VizPT is a free, browser-based tool that turns a short video into an annotated clip. Every moving subject gets a numbered mask labeled with what the subject is and what it is doing. You can download the annotated video or save any single frame as a PNG.

How it works

You choose a clip. Drag it in, pick from your library, or record live using your camera. The clip stays in your browser.
The AI reads the motion. Every subject visibly moving or interacting is identified, labeled, and given a numbered mask.
The overlay is drawn live. While the clip plays, masks mark their subjects. A side panel lists every subject the AI picked up.
Export. Hit Download Annotated Video for a shareable lightweight MP4 with the masks, numbers, and labels burned in, or Current Frame for a PNG of a single moment.

Privacy model

Your original video never leaves your browser. Only small downgraded image samples travel to our server for analysis. See the full Privacy Policy for the specifics.

Use cases

Journalists and analysts who need a fast structured read of user-submitted clips.
Product teams testing action-recognition workflows before writing custom code.
Coaches and instructors who want a per-second log of movement in a training clip.
Researchers benchmarking commercial vision models on their own footage.
Security and physical-space operators reviewing short incident clips.
Creators captioning and tagging short-form video.

What VizPT is not

VizPT is not a surveillance tool, a facial recognition system, a doxxing assistant, or evidence-grade software. AI output is statistical and sometimes wrong. Do not deploy it in systems where an incorrect label has legal, medical, or safety consequences. See the Terms of Service for the full list of disallowed use cases.

Related tools

PicPT — the image-analysis companion to VizPT. Drop a still photo, get bounding boxes, object labels, action verbs, scene narrative, and inferred subject thoughts. Built by the same operator.

Frequently asked questions

Is VizPT really free?

Yes. No signup, no credit card, no watermark on your own video file. A per-IP rate limit applies to keep the server running.

What formats are accepted?

MP4, MOV, and WebM. Maximum size 50 MB, maximum duration 20 seconds.

Does it work on mobile?

Yes. The Record button opens the in-browser camera recorder on mobile and desktop alike.

Why are some subjects missing?

The AI only places boxes on subjects it can clearly identify as moving. Weaker guesses are suppressed to keep the annotated output clean.

Can I use VizPT for commercial work?

Yes, within the Terms of Service. You retain ownership of your clip and the exported annotated video.

Is VizPT open source?

Not currently. The operator ships it as a live service; source access may be offered under a consulting engagement via sandeepkukreja.net.

Every Frame, Decoded

Record Video

Analysis Results