AI video restoration for old footage
Old video carries a mix of problems at once: low standard definition resolution, tape noise, soft focus from cheap lenses, and blocky compression from the digitizing step. This video restoration page is set up for that kind of material, with guidance and examples built around camcorder recordings, VHS and Hi8 transfers, and clips from the first phones that could film. The free AI video upscaler runs the same model for any footage and is the better fit when your clip is recent and simply small. Pick this page when the clip is old and you want it to look like a cleaner copy of itself.
Restoring photos from the same era? The free old photo restoration tool repairs scanned prints one image at a time.
How to restore old video with a free short-clip pass
The clip goes through a super-resolution model that upscales it while rebuilding edges and texture. RelayVSR, the default, keeps that detail steady from frame to frame; GAN is the quicker pick for longer clips. On old footage that does three useful things together. It sharpens soft outlines on faces, furniture, and landscapes. It smooths the fine noise and the blocky artifacts that tape transfers and old phone encoders leave behind. And it raises standard definition toward HD. The audio is copied over untouched, so voices and music stay exactly as recorded.
Prepare the clip first. The free tool takes up to 5 seconds with RelayVSR or 10 with GAN or Real-ESRGAN, 720p on the longest side, and 50 MB, which fits a typical 720 by 480 or 640 by 480 transfer. Cut the moment you care most about with the free video trimmer, upload it, and compare the result on the before/after slider.
How the tape was digitized matters more than most people expect. Capture at the tape's native resolution and skip any "enhance" or upscale option in the capture software, because those filters bake in smearing that the model then sharpens as if it were real detail. If you already have a file that was enlarged to 1080p during transfer, scale it back down to its original size with the free video resizer before restoring it. A clean, small source gives a better result than a large, blurry one.
What old video restoration can only partly fix
It helps to know where a general model stops. Scratches, dust specks, and dropout lines from a worn tape are sharpened along with everything else; the model sees them as part of the picture, so they become cleaner but they do not disappear. Faded or shifted color is kept as it is, because this pass restores detail rather than regrading the image. Interlaced footage, where moving edges show comb-like lines, should be deinterlaced during capture or export before you upload it. Tracking wobble and rolling bars from a damaged tape are also out of scope.
Expect a clearer version of the original, not a new film. Fine detail that the tape never recorded, such as the features of a face at the far end of a room, cannot be recovered reliably, and a restoration that claims otherwise is inventing it.
Make restored clips play smoother
Older recordings often feel jerky on a modern screen. For fluid motion, run the restored file through the free video frame interpolation tool, which generates in-between frames for smoother playback. Both steps are free, need no sign-up, and add no watermark. A small daily allowance of free GPU time covers several short clips, and connecting a Hugging Face account raises it.
Order matters when you combine the two. Restore first, then interpolate, so the generated frames are drawn from the cleaner, sharper picture rather than from the noisy original. Keep your original transfer as well: restoration is a copy, and the raw file is what any future tool will work from. For a whole tape, pick the moments that matter most, such as a birthday, a first step, or a wedding toast, restore those as short clips, and save the longer projects for the premium Video Upscaler, which restores clips up to five minutes in one job.