Upscale video in your browser with AI (Real-ESRGAN). The video edition of the Shaki image upscaler. No upload, completely free. MP4, MOV and WebM.
MP4 / MOV / WebM
Up to {n} px on the long edge · {s} s · 30 fps
Shaki Video v1.0 released. The same AI (Real-ESRGAN) as the Shaki image edition, applied to video at 2x.
Found a bug? Please tell us via the contact form.
Drag and drop the video you want to upscale, or click to choose it. MP4, MOV and WebM are supported.
Drag the two ends of the trim bar to set the range to export (up to 30 s on PC, 10 s on phones). "Play this range" shows what will be used.
The AI (Real-ESRGAN) analyses the video frame by frame in your browser and runs the upscale. The model (about 2.4 MB) is downloaded on the first run only.
Check the upscaled video in the preview and download it as an MP4. The audio is carried over untouched.
Yes, completely free. No account, login or sign-up is needed, and there is no limit on how many videos you process.
No, the video is never sent to a server. Everything, including the AI, runs in your browser (locally), so private videos are safe too.
Input: MP4, MOV (H.264, HEVC), WebM and anything else your browser can play. Output: H.264 MP4, with the original AAC audio carried over losslessly.
An open-source AI super-resolution technology called Real-ESRGAN. The same general-purpose model as the Shaki image edition is applied to every frame of the video, filling in detail while increasing the resolution.
Input up to 1280 px on the long edge (720p), 2x scale, output up to 2560 px. Larger sources are fitted to 1280 px first, then upscaled 2x. Length up to 30 seconds on PC and 10 seconds on phones, frame rate up to 30 fps. Full HD (1920 px) input is planned for the app editions.
Move the range and export in several parts, then join them in a video editor. Unlimited length and 4x are planned for the Windows edition (paid, coming soon).
Shaki Video needs a browser with WebGPU: desktop Chrome or Edge, or Safari 26 or later (iOS 26 or later on iPhone). If it does not run, close other tabs and try again.
The original audio (AAC) is placed into the new video without re-encoding, so the quality does not change. Unusual audio formats may not be carried over; in that case the result says "no audio".
Yes, on iPhone (Safari on iOS 26 or later) and Android (latest Chrome). The AI takes much longer than on a PC, so start with a short range; for long videos, a PC is recommended.
The AI model is the same as the image edition. Because video needs far more computation, the input is limited to 1280 px on the long edge, the scale is fixed at 2x and there are limits on length (30 seconds on PC, 10 seconds on phones) and frame rate (30 fps).
Processing time is decided by your device's GPU and the video resolution. As a guide, a gaming GPU takes a few minutes for 10 seconds of 1080p and under a minute for 480p or smaller; integrated graphics and phones take several times longer. Closing other tabs and apps and plugging in the charger help.
Videos processed with Shaki Video may be used freely, personally or commercially, by whoever holds the rights to the input video. The copyright of the original stays with its owner. This service claims no rights to your videos.
When a video is shown on a large screen the player interpolates between pixels to fill the gaps. That recovers nothing that was lost, so the bigger the picture the softer and more smeared it looks. Shaki Video uses super-resolution: an AI trained on huge numbers of blurry and sharp image pairs uses its knowledge of how textures and edges look to add pixels to every frame. Detail that the source did not contain is restored, within the limits of what still looks natural.
The best candidates are low-resolution, lightly compressed or simply older videos: footage from early smartphones and camcorders, clips that came back small from social media, screen recordings. Footage that is badly out of focus, crushed into blocks by extreme compression, or dark and noisy is harder for the AI to reconstruct and may improve less than hoped.
Redrawing the picture does not require touching the sound. Shaki Video extracts the audio data from the original file and places it, unchanged, into the new video. No re-encoding means the quality is identical and the picture and sound never drift apart. The audio of iPhone recordings (MOV) is handled the same way.
Shaki Video downloads its AI model (1 to 3 MB) once and then computes directly on your device's graphics hardware. Nothing is sent to a server, so family videos and work material are safe. The trade-off is that speed depends on your device: a gaming graphics card finishes in tens of seconds, laptop integrated graphics and phones can take a few minutes.
Most video upscaling services upload your video to a server and come bundled with an account, a queue and an upsell to a paid plan. Shaki Video differs on four points: (1) nothing is ever uploaded, it runs entirely in the browser, (2) no account and no install, (3) no watermark, unlimited and free, (4) the audio is carried over losslessly. The length and 30 fps limits of the free edition exist so that everything can finish on your own device; for 4x and unlimited length, a Windows edition is being prepared.