video2x-webGUI

A web interface for Video2X — AI upscaling (Real-ESRGAN, Real-CUGAN, Anime4K/libplacebo) and frame interpolation (RIFE) of videos on your GPUs — packaged as a linuxserver.io-style Docker container, so it drops into Unraid or any other Docker host.

Video2X's desktop GUI only exists for Windows; on Linux there's just the command line. This project puts a browser UI on top of the CLI, queues work across several GPUs, and runs whole folders unattended.

Unofficial — not affiliated with, endorsed by, or supported by K4YT3X or the Video2X project.

Features

• Every Video2X option in the browser — processor, model, scale, noise level, encoder settings. Only model/scale/noise combinations that actually work are offered.
• Batch folders — point it at a folder and every video in it gets processed. Outputs go into a matching folder (/batch/Show/x.mkv → /videos/output/Show/x.mkv), finished inputs are moved to done/, so an interrupted batch simply continues where it stopped.
• Multiple GPUs — each GPU runs one job at a time. Batch files wait in a shared pool and whichever GPU is free takes the next one, so a faster card simply does more of them. Single jobs you start meanwhile run next on their GPU.
• Preview clips — render a few seconds with the current settings and watch them right in the browser before committing hours to a whole season.
• Status page — live progress per GPU and per batch, time estimates, cancel anything, running/waiting counts in the tab title and browser notifications when a job or batch is done.
• Video info — resolution, frame rate, duration and the resulting output size, for single files and whole folders.
• Logs — one log file per job (with the exact command used), readable and deletable in the UI.

Batch mode	Preview clip

Requirements

• An x86-64 (amd64) Docker host.
• A Vulkan-capable GPU:
  • NVIDIA (best tested): current driver and the NVIDIA Container Toolkit on the host; the container gets the GPUs via --gpus all.
  • Intel / AMD (untested): pass /dev/dri into the container; the image includes Mesa's Vulkan drivers. Integrated GPUs, if they work, are far slower than a dedicated card.
  • Without any GPU, Mesa's software renderer (llvmpipe) still works — fine for trying the UI, far too slow for real work.

Tested on

• NVIDIA GeForce RTX 3090 and RTX 5060 Ti, on Unraid with the Nvidia-Driver plugin
• Mesa's software renderer (llvmpipe), for the automated build test

Other GPUs should work wherever Video2X itself works, but I can't test them. GPU-specific problems are almost always Video2X or driver issues — please report those to the Video2X project.

Quick start

Unraid

Search for video2x-webgui in Community Applications. Until it's listed there, add the template by hand: save unraid/video2x-webgui.xml as /boot/config/plugins/dockerMan/templates-user/my-video2x-webgui.xml on your server, then pick it under Docker → Add Container → Template.

The template sets PUID=99, PGID=100 and UMASK=000, but no GPU, so it starts on any server. Add your GPU by hand, otherwise Video2X runs on the CPU (extremely slow):

• NVIDIA: install the Nvidia-Driver plugin, then add --gpus all to Extra Parameters (advanced view).
• Intel/AMD: set the Intel/AMD GPU device to /dev/dri.

Both can be used together.

Docker Compose

services:
  video2x-webgui:
    image: ghcr.io/sognix/video2x-webgui:latest
    container_name: video2x-webgui
    environment:
      - PUID=1000
      - PGID=1000
    volumes:
      - ./config:/config
      - ./videos:/videos
      - ./batch:/batch
      - ./logs:/logs
    ports:
      - 3000:3000
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
    restart: unless-stopped

The same file is in docker-compose.yml. For Intel/AMD, replace the deploy: block with devices: ["/dev/dri:/dev/dri"].

docker run

docker run -d \
  --name video2x-webgui \
  --gpus all \
  -p 3000:3000 \
  -e PUID=1000 -e PGID=1000 \
  -v /path/to/config:/config \
  -v /path/to/videos:/videos \
  -v /path/to/batch:/batch \
  -v /path/to/logs:/logs \
  ghcr.io/sognix/video2x-webgui:latest

Then open http://<host>:3000.

Usage

• Single file: upload a video (lands in /videos/input) or copy it there, pick processor and GPU, start. The result goes to /videos/output.
• Batch: put a folder of videos under /batch, switch to Batch folder, pick the folder and the GPUs to use, start. Non-video files in the folder are ignored.
• Try settings first: open 6. Preview clip, render a few seconds and compare.
• Watch progress on the Status tab, read or clean up job logs on the Logs tab.

Configuration

Browser notifications need a secure page: https:// with a certificate the browser trusts (easiest via a reverse proxy) or localhost. The job counts in the tab title work everywhere.

Security

The web UI has no login. Anyone who can reach the port can upload files, start and cancel jobs and delete logs. Run it only on a trusted network, or put it behind a reverse proxy with authentication (e.g. Authelia, Authentik, or basic auth in nginx/Traefik/Caddy). Don't expose it to the internet as-is.

To report a vulnerability, see SECURITY.md.

Known limitations

• amd64 only — Video2X only ships an x86-64 Linux build.
• Hardware decoding (hwaccel) is unreliable in Video2X itself (upstream issue), so it's off by default. It rarely matters: the upscaling runs on the GPU either way.
• NVENC encoding (h264_nvenc / hevc_nvenc) currently fails with a pts < dts muxing error in Video2X; use a CPU encoder such as libx264 / libx265.
• Only the models bundled with Video2X are available. Supported combinations:
  • RIFE: rife-v4.6, rife-v4.25, rife-v4.25-lite, rife-v4.26
  • Real-ESRGAN: realesr-animevideov3 (2x/3x/4x); realesrgan-plus-anime and realesrgan-plus (4x only)
  • Real-CUGAN: models-nose (2x, noise 0); models-pro (2x/3x, noise 0 or 3); models-se (2x any noise, 3x/4x noise 0 or 3)
  • libplacebo: all 7 bundled shaders, any resolution
• The job list lives in memory: a container restart forgets queued jobs (finished files, logs and the done/ folders stay).

Documentation

• docs/architecture.md — how the image, queueing, preview clips and logs work
• docs/api.md — the HTTP API behind the UI
• CONTRIBUTING.md — building and running it yourself
• CHANGELOG.md — what changed in each release

Support

This is a hobby project, maintained in spare time. Bug reports with the details the issue template asks for are very welcome and get looked at; feature requests too, but without any promise. For problems with Video2X itself (models, output quality, crashes of the video2x binary), please check the Video2X repository first.

License

Copyright (C) 2026 sognix — AGPL-3.0-or-later, see LICENSE.

The image bundles Video2X (AGPL-3.0), FFmpeg from Ubuntu (GPL-2.0-or-later) and the linuxserver.io base image; see NOTICE.md for all third-party components and their licenses.