How I Turned Blurry Videos into Clearer Footage with ComfyUI

’ve built an open-source ComfyUI custom node package called ComfyUI MediaKit Toolkit.

The idea is simple: generate at a lower resolution to save on production costs, then use MediaKit to enhance the final video. This “low-resolution generation, high-quality delivery” workflow helps reduce token and compute consumption without giving up a clear final result.

What it does

The plugin connects ComfyUI to Volcengine AI MediaKit and exposes its video-processing capabilities as native ComfyUI nodes.

Currently available:

  • Enhance · 视频增强 — standard and professional video enhancement

  • Enhance Ultra · 视频增强(大模型版) — generative, large-model video restoration

  • Erase Subtitle · 字幕擦除 — automatic hard-subtitle removal

  • Erase Subtitle Pro · 字幕擦除(增强版) — advanced subtitle and rendered-text removal

  • MediaKit Environment Check — CLI, authentication, and environment diagnostics

Native video workflow

The nodes use ComfyUI’s native VIDEO input and output:

Load Video
    ↓
Enhance / Erase Subtitle
    ↓
Video understanding, OCR, generation, or other nodes
    ↓
Save Video

You can select a local video in ComfyUI. The plugin automatically handles:

  • Local video materialization

  • Upload to MediaKit

  • Cloud task submission

  • Asynchronous status polling

  • Result download

  • Redacted diagnostic logs

  • Native VIDEO → VIDEO workflow integration

There is no need to manually upload a file on another website, wait for processing, download it, and import it back into ComfyUI.

Project

GitHub:

The repository includes Chinese and English documentation, example workflows, installation instructions, API-key setup, and troubleshooting guidance.

The automated test suite currently passes on Windows, macOS, and Ubuntu with Python 3.10 and 3.12.