YouTubers are using AI software like DeOldify to upscale and “enhance” historic footage into 4K, but historians say the result undermines the footage
Thomas Nicholson / WIRED UK : Tweets: @wireduk , @samokhvalov , @chastaylor , @citnaj , and @jcalpickard Tweets: @wireduk : Digital upscalers and the millions who've watched their work on YouTube say they're making the past relatable for viewers in 2020. But for some historians of art and image-making, modernising century-old archives brings a host of problems https://www.wired.co.uk/... Nikolay Samokhvalov / @samokhvalov : Video archeology powered by ML in hands of a guy with a Russian accent. [60 fps, 4K] The oldest recorded video, England,1888 https://www.youtube.com/... “YouTubers are upscaling the past to 4K, but historians want them to stop” https://www.wired.co.uk/... HN: https://news.ycombinator.com/ ... https://twitter.com/... Charlie Taylor / @chastaylor : Interesting read... https://www.wired.co.uk/... Jason Antic / @citnaj : 1/ My cofounder @danasday and I were interviewed by Wired, and quite enjoyed it. It's about the controversy of using Deoldify colorization and other AI techniques to enhance photos and video from the past. https://www.wired.co.uk/... Justin Pickard / @jcalpickard : Interesting profile of a Polish firm using machine learning to buff and upscale old footage, unsettling historians. https://www.wired.co.uk/...
Context & Ripple Effects
This WIRED UK piece lands mid-arc in a 2020 wave of AI video restoration coverage: weeks earlier, Wired profiled Neural Love, whose AI-powered tools upscale, colorize, and sharpen old films to 4K at 60 fps. The DeOldify story extends that from tooling to audience — millions watching YouTubers' enhanced archives — and adds the counterweight: art and image-making historians who argue modernizing century-old footage undermines it.
The dispute is an early instance of a question that has since hardened on YouTube itself, where the platform went on to terminate large channels for AI-fabricated trailers and a Kapwing study found top channels posting nothing but AI-generated video. What was framed in 2020 as enhancement-versus-authenticity is now part of the platform's synthetic-content moderation problem.
First-order effects
- Upscaling channels gain reach by making the past 'relatable' to 2020 viewers, while historians of image-making lose interpretive ground: interpolated frames, invented color, and sharpened detail present fabricated pixels as historical evidence.
- Toolmakers like DeOldify's Jason Antic and Neural Love see consumer demand validated — their software is now the visible engine of a popular YouTube genre.
Second-order effects
- Archives and rights-holders of historic footage face pressure to assert provenance and license terms, since third-party 'enhancements' can outperform and overshadow the originals they derive from.
- YouTube inherits an authenticity-moderation burden that foreshadows its later crackdowns on AI-fabricated content, including the termination of Screen Culture and KH Studio for fake movie trailers.
Third-order effects
- If enhancement keeps outrunning disclosure, historical media splits into two tiers — verified originals and AI-reconstructed derivatives — echoing researchers' warnings that neural networks can slip bias or falsifications into historical records.
- The pattern points toward platform-level labeling norms for AI-altered media, the same policy terrain YouTube entered when policing synthetic 'slop' channels that the Kapwing study found amassing tens of billions of views.
The trend: Generative enhancement is turning historical archives into engagement-optimized YouTube content, forcing a widening split between restoration and fabrication that platforms and archivists are still learning to police.