Welcome to DeepGuard, your second pair of eyes online! Whether you’re fact-checking something serious or unsure about a viral clip, just upload or paste a link to get a clear, thoughtful verdict. Spot deepfakes and understand what’s real or not, fast.
DeepGuard
DeepGuard is super useful and straight to the point. With all the deepfakes out there, this feels like a must-have. Would love to see a browser extension in the future.
Great job! 🙌
minimalist phone: creating folders
This directly travelled to my collection "interesting" in the Product Hunt bookmarks. BTW, which tools do you think create the most authentic AI videos?
P.S. I tried to upload the video, but the button doesn't react :/
+ one suggestion on the landing page. I would personally use the different CTA instead of launching the app. E.g. "test video"
DeepGuard
@busmark_w_nika Thank you for adding it to your collection! 🙌 There are generally two kinds of deepfakes we’re seeing out there, the visually impressive ones , like Sora, Pika, and Runway! They’re amazing at generating cinematic realism, but usually easy to spot if you’re paying attention. On the other hand, the more subtle ones, often created with tools like DeepFaceLab or by editing real footage, are much harder to detect. That’s exactly the kind we built DeepGuard to catch :)
P.S You’re totally right, there’s no upload function just yet. For now, you can just click on any of the sample videos and run a test that way!
Also, really appreciate the CTA tip! We’ll take a closer look at that.
Augmented Reality 2021
How accurate is this? And how do you know?
DeepGuard
@danischenker Great question! DeepGuard analyzes both the video itself and related web context. On the video side, we use multiple agents to detect things like motion inconsistencies, visual artifacts, and manipulation patterns. At the same time, we gather signals from the web, such as where the video has appeared, how it’s being used, and whether similar versions exist. The final verdict combines both layers for a more reliable result!
Augmented Reality 2021
@lucianb Thanks for the insights. Although, you didn't really answer my question. Of course, you could just say that you don't have an answer ;-) I was looking for a percentage. Or maybe, how many false positives you get... or what ever...
DeepGuard
@danischenker Right now, we don’t publicly show a fixed accuracy % because DeepGuard works across multiple layers, pure video analysis and contextual web signals, and the outcome can vary depending on the input :)
That said, in internal tests, DeepGuard has shown around 87–92% accuracy across a wide range of deepfake samples, especially those involving subtle manipulations that are hard to catch by eye. We’re continuing to refine and stress-test it, and by July 17, we’ll be sharing more detailed benchmark data, including false positive/negative rates! Stay tuned :)
Congratulations on the launch! We definitely need a tool like this nowadays.. AI video generators getting so realistic, it’s becoming almost impossible to tell what’s real and what’s not unless there’s a clear label.
Really love what you’re building here. Excited to see where it goes 👏
DeepGuard
@isibol01 Thank you Ilkkan! 🙌 Totally agree, it’s getting harder to tell what’s real out there. We’re building DeepGuard to help bring some clarity back. Really appreciate the support! 💚
Mozart AI
Seems really interesting! I’m curious how the system handles low-quality videos, like those with heavy compression, low light, or older camera tech. Do you apply any preprocessing to reduce false positives in those cases?
Also, a browser extension that checks videos in real-time (e.g., on social media or YouTube) could be a game-changer. Is something like that on your roadmap?
DeepGuard
@petr_ivan Appreciate your comment! You’re highlighting two big areas we’ve been working on behind the scenes 🙃
Handling low-quality videos is definitely a challenge. Right now, DeepGuard applies some light preprocessing to normalize and clean up frames, but what really helps is our multi-agent system, each agent looks at different angles: visual patterns, sync, motion, metadata, and even web signals. This layered setup helps reduce false positives, especially on compressed or older footage.
And about the browser extension, spot on! We actually built a working prototype already, and it’s in the pipeline for release. The idea is exactly what you said: bring detection closer to where people see content in real-time (social media, YouTube, news sites). Can’t wait to ship that one!
Thanks again Petr!
Why did you choose to launch DeepGuard with only sample videos and no upload or sharing features, and how soon will users be able to test it with their own real-world content?
DeepGuard
@vouchy We launched with sample videos first to ensure a consistent experience across all users while we fine-tune detection and analysis at scale. It also allows us to showcase the system’s core capabilities without dependency on upload quality or unsupported formats
Support for uploading real-world content and sharing results is already in place and will go live with our full launch on July 17. Early users who explore the demo now will be first in line, with a few extra perks coming their way!