PVPrivacy policy
Your video stays yours.
This page says what happens to your video and your data when you use Frame Recon, in plain words.
Last updated 14 November 2026
- Your video file is never uploaded. It is opened and scanned by your own browser.
- Asking a question sends still frames, not the video. They go to Google's Gemini model to be read, and we keep no copy.
- We have no database. There is no account on our side and no record of what you scanned or asked.
- No adverts, no analytics, no tracking.
Who runs Frame Recon
Frame Recon is an independent website. A contact address will be listed here before the site opens to the public.
What stays on your device
The scan runs in your browser, and so does everything it produces.
- The video. Your browser reads the file from your device. It is not sent to us or to anyone else.
- The scan. The scenes, the objects found, the transcript and small preview pictures are saved in your browser's own storage, so you can come back to a scan later.
- A private copy of the video. For files up to 1.5 GB, the browser keeps its own copy so a scan can carry on after a refresh. For bigger files, Chrome and Edge remember your permission to reopen the original file.
- The AI models that do the scanning. They are downloaded once and kept on your device.
To remove a scan and its copy of the video, use Delete in the list of earlier scans on the scanner page. Clearing this site's data in your browser removes everything at once.
Frame Recon never uses your microphone or your camera. The live scan of a YouTube video watches a browser tab that you choose to share, without its sound, and those pictures are analysed on your device.
What is sent to the AI model
Only what a question needs, and only when you ask.
Three features use Google's Gemini model, and all three need you to be signed in. Nothing is sent until you use one of them.
- Ask, for a video file. Your question, a written summary of the scan (the scenes, the objects found and what was said, with their times), and the still frames the model asks to look at. Your browser cuts those frames from the video. The earlier questions and answers in the same conversation are sent again with each new question, because our server keeps none of them.
- Deep scan. One still frame from each scene, up to 48 frames a run.
- A YouTube link. The link and your questions. Google's model reads the video from YouTube itself.
These pass through our server on the way to Google and the answer comes back the same way. Our server does not save them to a disk or a database. If a request to Google fails, the server logs the error message it got back so the fault can be fixed.
Every request is sent with the setting that tells Google not to store the exchange. Google handles what it receives under its own terms for the Gemini API, which you can read in the Gemini API terms.
Signing in with Google
Sign-in exists to stop the AI features being abused. There is no profile behind it.
When you sign in, Google tells us your name, your email address, your profile picture and an account number. We keep the account number, the name and the picture in an encrypted cookie in your own browser, for seven days. We do not keep your email address, and we do not write any of this to a database, because there is none.
We use what Google sends only to sign you in and to count your daily use of the AI features. We do not sell it, share it or use it for advertising.
Signing out removes the cookie. You can also take away Frame Recon's access from your Google account at any time, on Google's third-party connections page.
Cookies and browser storage
Only the ones the site needs to work.
- Sign-in cookies. One holds your session, as described above. A few short-lived ones protect the sign-in step itself.
- A usage counter. A signed cookie that counts how many AI requests you have made today. It lasts about a day.
- Browser storage. Your scans, the private video copies and the model files, as described above. It also holds the summaries of YouTube videos you have opened, and a note that you have seen the live scan tips.
There are no advertising cookies and no analytics cookies.
Other services your browser talks to
- Hugging Face. The AI models that run on your device are downloaded from huggingface.co the first time they are needed. Like any web server, it sees your IP address and browser type. Nothing about your video is sent there.
- YouTube. The YouTube page plays videos in YouTube's own player, in its privacy-enhanced mode. Playing a video there is covered by Google's privacy policy.
- Our host. The site is served by Vercel, which keeps ordinary server logs such as IP address, time and the page asked for.
What we do not do
- We do not sell or share your data.
- We do not show adverts or track you across sites.
- We do not use your videos, frames or questions to train models of our own.
- The scanner does not recognise faces and will not tell you who someone is.
Your rights
Depending on where you live, the law may give you rights over your personal data, such as seeing it, correcting it or having it deleted. We hold almost nothing: no account, no history, and no copy of your videos. What exists is on your own device, where you can delete it yourself. If you have a question or a request, get in touch. A contact address will be listed here before the site opens to the public.
The AI features need a Google account, so they are only for people old enough to have one.
Changes to this page
If the way Frame Recon handles data changes, this page changes with it and the date at the top is updated.