Nodes/ComfyUI-SA-Nodes-QQ/Power Load Video (QQ)
ComfyUI Node

Power Load Video (QQ)

Load videos in ComfyUI with a built-in timeline player

By siraxe·Created 11 months ago·Updated a day ago· 72
Power Load Video (QQ)
  • metadata
  • image
  • high_images
  • audio
  • frame_count
  • metadata
video
start_frame1
end_frame-1
force_fps0
max_fps0
crop_enabledfalse
crop_x0.50
crop_y0.50
crop_w1.00
crop_h1.00
width0
height0
high_size1.0

The stock ComfyUI video loaders are fine for "load file, get frames," but they give you nothing to look at. PowerLoadVideo is the version that treats video like video: a drag-and-drop file loader that previews in-canvas with an actual timeline - play/pause, loop boundaries, the works. When you're building a Wan workflow around an existing clip, that preview isn't a luxury, it's the thing that lets you find the frame you actually want without exporting to an external player.

It's one of the flagship nodes of siraxe/ComfyUI-WanVideoWrapper_QQ, sir_axe's experimental Wan pack - and it's arguably the pack's best advertisement. The README calls out that it can override ComfyUI's default drag-and-drop behavior so a dropped video creates this node instead of a plain loader, but that's optional and off by default.

How it works

Under the hood it's OpenCV doing the heavy lifting (cv2.VideoCapture), reading frame-by-frame into an IMAGE tensor of shape [frame_count, H, W, 3] while also extracting the audio stream into ComfyUI's AUDIO format (a waveform/sample-rate dict). The FPS is read from the container's metadata. Nothing exotic - the value is the UI wrapped around it: the preview has a timeline with [ / ] loop markers you can scrub, Shift+Spacebar to play/pause, and Shift+Drag on the markers to set loop boundaries.

Inputs and outputs that matter

Everything's optional, which is the point - it's a loader, so most of these are conveniences:

  • video - the file picker. This is also where you drag-and-drop a video.
  • start_frame / end_frame - trim range (1-based; -1 = run to the last frame).
  • force_fps - reinterpret the output FPS, same logic as VideoHelperSuite's force_rate. 0 = keep native.
  • max_fps - cap the output frame count by sampling the source at a lower rate. Handy when you need a preview-speed pass before committing to a full render.
  • crop_enabled + crop_x/y/w/h - normalized (0–1) crop before output, useful for squaring up a 16:9 clip for a 1:1 Wan generation.

Outputs are IMAGE, AUDIO, and FPS. The FPS output is the sleeper: wire it into a Wan workflow's video-combine or sampler setup so your output video honors the source's frame rate instead of assuming 24.

Installation

cd ComfyUI/custom_nodes
git clone https://github.com/siraxe/ComfyUI-WanVideoWrapper_QQ.git

Restart ComfyUI, or grab it via ComfyUI Manager (search "WanVideoWrapper_QQ"). Two practical notes: this node needs OpenCV (opencv-python), which the pack's empty requirements.txt won't install for you - if cv2 isn't already in your ComfyUI env, pip install opencv-python in that environment. And if you've got an old wanwrapper_qq folder from before the v1.3.4 rename to ComfyUI-SA-Nodes-QQ, delete it to avoid import collisions.

Common issues

The big one: on the very first run after enabling it, the drag-and-drop override can feel like ComfyUI "broke" - if you didn't want that behavior, the README notes it's optional and off by default, so check you're on a recent version before blaming your install. Otherwise, expect the classic OpenCV quirks: some codecs (HEVC in particular) will fail silently depending on your ffmpeg/opencv build, and exotic frame rates read back as odd values. The crop sliders are normalized, so remember crop_w: 0.5 means "half the width," not "500 pixels" - a common first-run surprise.

CategoryPower/Video

Inputs (14)

NameTypeDefaultDescription
videooptCOMBO0 options:
start_frameoptINT1
end_frameoptINT-1
force_fpsoptFLOAT00–60
max_fpsoptFLOAT0
crop_enabledoptBOOLEANfalse
crop_xoptFLOAT0.500–1
crop_yoptFLOAT0.500–1
crop_woptFLOAT1.000.05–1
crop_hoptFLOAT1.000.05–1
widthoptINT0
heightoptINT0
high_sizeoptFLOAT1.0
metadataoptMETADATA

Outputs (5)

NameTypeDescription
imageIMAGE
high_imagesIMAGE
audioAUDIO
frame_countINT
metadataMETADATA