AI video is getting better and more common
Face swaps, cloned voices, and generated scenes now blend into normal social feeds. As image and video models improve, small cheating becomes easier and harder to spot.
Prove your footage is Real.
As AI-generated media becomes indistinguishable from reality, RealRec lets you prove your photos and videos were captured from a camera.
Why this is needed
Social clips are hard to judge as real or fake
Viewers need a reliable way to check authenticity
Trust is social currency. A physique photo, a skate trick clip, or a clip of a politician only has value if people can know it is real. If viewers suspect AI fabrication, the post loses meaning, so quick access to the original source becomes critical.
How verification works
How RealRec links shared posts to the original source
Every step keeps a direct reference from shared media back to the immutable source, so viewers can compare what changed and what did not.
- 1
Record only inside RealRec
You can capture media in the app, but you can never import existing files.
- 2
Original source is locked
Because capture is contained in the app, RealRec stores the true original and prevents user manipulation of that source file.
- 3
Export embeds source reference, source goes live
When exporting, RealRec overlays a small watermark with a QR code. At the same time, it uploads the original capture to the source server.
- 4
Viewer scans QR to open source page
On social media, people can screenshot the content and scan the QR to open the URL with the original source and capture properties. Not everyone needs to check — if even a few curious viewers scan and spot something suspicious, they can flag it in the comments for everyone else to see.
Practical for creators
Creators can still color-grade, trim, and polish after export in the regular photo library. The shared version does not need to be identical to the stored source to still depict real events. Viewers can compare both and use human judgment to decide whether edits changed authenticity.
Proof stays attached to media
The QR watermark carries the source location directly in the shared image or video. Saved copies and screenshots keep this info too, so people do not need a post description to find proof.
What does the QR-watermark look like?
Look at these examples, they all encode the link to their immutable source. In these the QR-watermark is in the bottom-left corner.
Hard edge case
If someone records fake content on a screen
This is the classic limitation of recording verification: the recording can be real, while what it depicts is fake. RealRec uses two capture-time tests to make this very hard to hide.
Zoom test (min and max zoom)
For every capture, RealRec takes two verification images at 0.5x and 5x (or min and max supported zoom). Screen setups are hard to hide at both levels. At wide zoom, environment or display borders often appear. At high zoom, pixels and display artifacts often appear.
Flash test (~50ms at video start)
The device flashes briefly at the beginning of video capture. Filmed screens typically show glare and reflections from this flash, while real scenes behave differently.
The recording looks like a real capture of the politician, but the Zoom-test reveals that it is actually a recording of a screen.
All the verifications are performed automatically when you take an image or video. You don't have to perform zoom or flash tests manually. The app can be used like a regular camera app.
Combined protection
Together, the Flash-test and Zoom-test make it significatly more difficult to record a screen without getting caught. To make a successful screen recording, the pixels or moire patterns cannot show in the 5x test, the boarders of the screen cannot show in the 0.5x test, the perspsective in the scene needs to transform correctly between the focal lengths of 5x and 0.5x. If the camera is close to the screen it will glare by the initial flash. This together doesn't make it 100% fraud proof, but it makes it much harder to produce dishonest screen recordings.
These tests are intentionally primitive and phone-friendly. They use common camera capabilities, so no LiDAR is required. The Flash-test can be turned off in situations where an initial flash would be unwanted. The Zoom-test can never be turned off.
Trust model
Why trust RealRec with verification?
Verification only works if people trust the verifier. RealRec is intentionally not designed as a zero-trust system. Focus is instead on open scrutiny, centralized trust and solutions that make manipulation difficult.
The problem with cryptographic zero-trust
One way to achive zero-trust in theory is cryptographic solutions. The problem is that when you upload your image/video to a platform, the media is often per definition not the same anymore due to encoding/compression, so you lose the cryptographic connection. This is also the case when slighly modifying it in a good faith way (colors, contrast, slight cropping etc). This makes it in practice difficult for cryptographic solutions to survive real world sharing.
Zero-trust outside the theoretic realm isn't really possible
Even if a system is mathematically designed to exclude trust, when it is implemented in practice, the user still has to trust the implementor. So the bottleneck is mostly moved, not eliminated.
Responsibility is centralized, not scattered
Many systems split trust across vendors and services. That means more handoffs, more weak links, and less clear ownership when something fails. RealRec keeps the critical path with one accountable team.
RealRec's state-chain
Every hour, a snapshot of the database ("resources" table) is uploaded to this website. Since frequent historic snapshots are availible for download/scrutiny, if we would change anything in bad faith, it's much harder to not be caught. The periodic snapshots lock in history and therefore also possible future trajectories of the data.
Trust initiatives
External technical reviews
RealRec is open to independent review of capture integrity, source immutability, and verification-link behavior. External scrutiny is a direct way to increase confidence in implementation quality.
Open source, with hardening first
Open source can increase trust, but it also exposes implementation details to everyone, including attackers. Public code makes it easier to study potential weak points. That is why we harden first, then consider open-sourcing the app and website after external reviews give high confidence in security.
Interested in reviewing RealRec or have questions about our trust model?
hugo@hoverest.comFAQ
Questions one might ask
Short answers for specific details about the RealRec system that aren't obvious
Yes! RealRec is completely free. However, if RealRec becomes popular, then there will need to be some kind of small subscription for heavy users since the storage of all the immutable orignals and verification data costs money.
Yes! You can always export images or videos you have captured without verification or a QR-watermark. The exported image will then of course not be verified, but instead just a regular image/video.
Nothing harmful will happen. If the watermark is removed, then the link to the immutable original is gone. This just makes it to a normal image/video. If the removal was unintentional because a social media app cropped your image/video, then you can reexport your footage again with the watermark placed a bit differently so it doesn't get cropped away.
Not necessarily. Creators can still crop, trim, color-grade, or compress an export before sharing it. This is an important feature to preserve. The QR watermark points viewers back to the original source so viewers can compare what changed. This allows for good faith changes while still not losing the verification system. Verification means that it was captured from a camera and that an reference to its immutable original and zoom/flash test is embedded.
They are redirected to a page with the original capture. There they can see verification information, zoom/flash-tests, export history and more to judge if the media is real or not.
No! RealRec will not allow any imports, so anything that can come from RealRec must be captured by the camera inside the app.
No. The watermark is very small and half transparent. You can also position it wherever you want with the export-settings in the app. The QR-watermark is designed to be the smallest possible while still have enough pixels to be scannable (even after social media compression).
Yes! It is possible to export as verified without the watermark. For some cases where it is very important to not have a watermark at all, even if it is tiny, you can export and get a small code instead. When you then publish you image/video on for example Instagram, you just put the code in the post-text.
When verifiying without an embedded QR-watermark, you get a code that you can publish together with the image/video. This gives enough information to link the image/video to the immutable source. The advantage of this is that you don't get a small watermark, the disadvantage is that the immutable reference is no longer embedded in the media. This means that if someone republishes your image/video without the code, or that someone takes a screenshot, the verification might not survive.
Not yet. The first version to test the waters is made for iOS. If the app gets used, then an Android version will be published.
Coming soon
iOS app
Get RealRec on the App Store
The iOS app is not published yet, but it is on its way. Soon you will be able to record with verification, export with an embedded source reference, and share with a traceable original.
Not published yet