AI Video Upscaler
100% free
No upload — on-device
No watermark
AI on-device
Run every frame through a real super-resolution network for a 2x upscale with recovered detail — not just bilinear stretching. Heavy by nature, so short clips only.
Real super-resolution
A neural network hallucinates plausible detail instead of just stretching pixels.
Honest limits
Per-frame inference takes around a second, so runs are capped at 8 seconds and 960px input. Trim first, upscale the moment that matters.
On-device
Frames are processed in WebAssembly on your machine.
Frequently asked questions
How long does it take?
Roughly 0.5-1.5 seconds per frame depending on your hardware - a 5-second clip at 30 fps takes a couple of minutes.
Why only 2x?
2x keeps quality high and time sane in the browser. Run the output again for 4x on very short clips.
Is audio kept?
Yes, the audio track passes through untouched.