Enhance · 视频增强

Let's be honest about what this node is: it's the "I'd rather pay than wait" option for video enhancement. Local video restoration is the least settled problem in the upscaling world - good temporal upscalers need datacenter-class GPUs (a 720p→4K SeedVR2 pass can take 12 minutes on a rented L40S, per the KB). This node offloads the whole job to Volcengine's AI MediaKit, which runs a standard or professional enhancement pipeline in the cloud and hands back a finished VIDEO. You don't render anything; you just get billed.

The inputs that matter

• tool_version - standard or professional. The professional tier is the more capable (and presumably pricier) pipeline. Try standard first; reach for professional when it's not enough.
• scene - common, ugc, short_series (short drama), aigc, or old_film. This tunes the enhancement to the content's look. Note it only applies to the standard version - the README says professional doesn't use it.
• resolution - keep, or pick from 240p up to 4k. keep preserves the source resolution; the rest are target resolutions for the upscale.
• bitrate_level - low, medium, high target encode bitrate. Medium is a sane default.
• fps - 0 keeps the source frame rate; set up to 120 to change it.

Plus video and the poll controls - this one polls faster by default (poll_interval_seconds default 5) with the same 720-attempt ceiling. The single output is video, straight into Save Video: Load Video → Enhance → Save Video.

How it works

Standard cloud choreography, which you've seen in every node in this pack: the video materializes to a temp file, mediakit-cli submits an async enhance task to your MediaKit account, the node polls on the interval, downloads the finished clip, and converts it back to ComfyUI's native VIDEO object. Nothing runs on your GPU. Everything bills your account and uploads your footage to Volcengine.

Install and the cloud setup

ComfyUI Manager, search "MediaKit Toolkit" / mediakit-toolkit, full restart - or:

cd ComfyUI/custom_nodes
git clone https://github.com/fudanll-sys/comfyui-mediakit-toolkit.git

No Python dependencies, no model downloads. You need a Volcengine account with AI MediaKit activated, an API key, and the CLI installed and authed in the environment that starts ComfyUI:

npx @volcengine/mediakit-cli install -y
mediakit-cli init --mode cloud-first --api-key "YOUR_MEDIAKIT_API_KEY" --credential-store config --yes

Where people get tripped up

Two things. First, this is not the fast edition - the README says so explicitly, because an ultra-fast enhancement capability exists but isn't wired in yet and will get its own node when it lands. Don't go looking for a speed setting here. Second, if what you actually want is a generative model that invents missing detail rather than a classic enhancement that cleans up, that's the separate Enhance Ultra node. And if you have the hardware and the patience, local options like SeedVR2 or RTX video super resolution are free - the cloud only wins on turnaround and convenience. For a one-off dirty clip with a deadline, though, this is hard to beat: send it, wait, get pixels back.