A look at Neural Love, which develops AI-powered video enhancement tools that can upscale, colorize, and sharpen old movies, bumping them up to 4K and 60 fps
Matt Simon / Wired : Tweets: @sammyandrews , @jenlucpiquant , and @gadgetlab Tweets: Sammy Andrews / @sammyandrews : This is beautiful but also very unnerving to watch 😂🤦♀ ️ People and horses running in front of trams & carts galore. But a lovely slice of history restored. https://twitter.com/... Jennifer Ouellette / @jenlucpiquant : AI Magic Makes Century-Old Films Look New. Denis Shiryaev uses algorithms to colorize and sharpen old movies, bumping them up to a smooth 60 frames per second. The result is a stunning glimpse at the past. https://www.wired.com/... @gadgetlab : Using AI, Denis Shiryaev has transformed a short 1906 clip of San Francisco into something magical. He colorized and sharpened the film to 4K resolution and bumped the choppy frame rate up to 60 frames per second. The resulting movie is mesmerizing. https://www.wired.com/...
Context & Ripple Effects
Denis Shiryaev's AI-restored 1906 San Francisco street scene — colorized, sharpened, and bumped to 4K at 60 fps using Neural Love's tools — is the viral proof-of-concept for a hobby that was about to get crowded: weeks later, YouTubers were running DeOldify and similar software over historic footage, drawing pushback from historians who argue the 'enhancement' distorts the record.
The Wired piece matters because it captures the moment before the tooling commoditized — three years on, Adobe was detailing Project Res-Up, an experimental upscaler that lifted a 1947 film from 480x360 to 1280x960, moving what Shiryaev hand-ran into mainstream creative software.
First-order effects
- Shiryaev and Neural Love get the visibility that comes with a viral restoration, positioning their enhancement tools as the reference workflow for archival upscaling while historians publicly question whether the output still counts as historical footage.
- Archives and rights holders of old film suddenly face demand — and pressure — to deliver 4K/60fps versions of material that was never shot anywhere near that fidelity.
Second-order effects
- Adobe's entry with Project Res-Up turns a niche enthusiast pipeline into a checkbox feature, forcing standalone tools like Neural Love to compete on quality and integration rather than on simply being the only option.
- The historian critique pushes platforms and archives toward labeling enhanced footage, since audiences can no longer assume a crisp, smooth video is what the camera originally captured.
Third-order effects
- If enhancement keeps migrating into standard editing suites, 'restoration' and 'reinterpretation' blur structurally: the archival record becomes a rendered output whose fidelity reflects an algorithm's assumptions about faces, motion, and color rather than the original emulsion.
- The same toolchain logic — AI filling in detail the source never contained — extends from old film to new footage, feeding the broader question, visible in Hollywood's adoption debates, of whether AI remains an assistant to creatives or starts defining what 'authentic' video means.
The trend: AI video enhancement is moving from viral one-off experiments into mainstream creative toolchains, dragging an authenticity debate about archival truth along with it.