4K Video Save Load - Hallett

ComfyUI 4K Faster Video Loader and Saver with Fun Features

Similar to the VideoHelperSuite video nodes, but with the file's information exposed on the node and much faster load and save for large files. Two nodes that do their per-pixel work on the GPU, so 4K stops being the slow part of a workflow.

Independent of VideoHelperSuite: it is not a fork and shares no code with it, so both can be installed at once.

<img width="344" height="1085" alt="4K_Load_Screencap" src="https://github.com/user-attachments/assets/989aaa4a-3030-4026-8521-a1860ee8ab19" />

Speed

Measured on an RTX 5090 at 3840x2160 with an idle GPU, milliseconds per frame. Load is 100 frames, best of two runs; save is 60 frames.

| | VideoHelperSuite | Video 4K | | | --- | --- | --- | --- | | Load, full resolution | 377.9 | 20.1 | 18.8x | | Load, scaled to 1080p | 45.6 | 9.0 | 5.1x | | Save, h264 nvenc | 87.8 | 30.9 | 2.8x | | Save, h264 cpu | 104.2 | 39.4 | 2.6x |

VHS's OpenCV loader reads full-resolution 4K at 43.5 ms/frame, so the ffmpeg path here is faster than either of the loaders it replaces. Measure with an idle GPU: a render in the background inflates these several times over.

Load Video 4K

One node in place of VHS's four loaders.

• source — upload uses the picker, path takes a file or a folder. For a folder, video_index chooses which clip and the file details show how many there are, so a queue can walk a whole directory.

• Drag and drop — drop a clip anywhere on the node, preview included. Drops and the picker take .mp4, .mov, .mkv, .webm, .avi, .m4v and .gif.

• resolution — source, 2160p, 1440p, 1080p, 720p, 480p, 2048 wide, 1920 wide, 1024 wide, or custom. Each preset sets the longest side (2160p is 3840, 1080p is 1920), so portrait clips get the same class of size as landscape ones. Aspect ratio is kept. Picking one fills in custom_width / custom_height with the size the clip will load at; typing in either switches to custom, and leaving one at 0 makes it follow the aspect. Scaling happens inside ffmpeg, which is far faster than resizing after decode.

• divisible_by — rounds both dimensions. The model preset raises it if needed.

• model_preset — sets the frame rate, the size rounding, and snaps the frame count to what the model actually accepts:

  | preset | fps | frame counts | size multiple | | --- | --- | --- | --- | | MiniMax-H3 | 24 | 17n + 5 (5, 22, 39, 56 …) | 32 | | LTXV | 24 | 8n + 1 | 32 | | Wan | 16 | 4n + 1 | 16 | | Hunyuan | 24 | 4n + 1 | 16 |

  As in VHS, the preset also makes frame_load_cap step through those counts: with H3 its arrows go 5, 22, 39, 56 and a typed value snaps to the nearest one.

• seconds_cap, frame_load_cap — show the length of the whole clip until you change one, and the other follows it. Combined with a model preset the frame maths is done for you: 3 s of H3 becomes 56 frames, not 72. ↺ goes back to the whole clip.

• skip_first_seconds, select_every_nth — as usual.

• force_frame_rate — whole numbers only. 0 keeps the source rate.

• audio_when_missing — silence emits a silent track matching the clip length so downstream nodes never break on a file with no audio. none outputs nothing.

When a widget's own value is not what will load, the real value shows beside it in grey, recomputed live from every other setting: ask for 158 frames of a clip that has 141 and frame_load_cap reads 141← 158. The ↺ beside the size, rate and cap widgets puts the clip's own value back, and is dimmed when there is nothing to reset. reset all to defaults does every setting at once and keeps the chosen clip.

The file's own details also appear in grey at the bottom of the node as soon as you pick it, before running anything, with the part that will load: loads 1.5s to 4.2s (2.7s, 65 frames). Picking, uploading or dropping a new clip resets the settings to their defaults; opening a saved workflow does not.

The preview plays only that part, looping it, and jumps to the new start or end as you change skip_first_seconds or a cap. Switch preview to whole clip to scrub the entire file, then set start here and set end here mark the part to load from where the preview is.

Outputs: images, audio, frame_count, fps, width, height, info (a JSON summary of source and loaded properties) and video, the same frames and audio as a VIDEO for nodes that take one. fps already accounts for force_frame_rate, the model preset and select_every_nth, so it can be wired straight into Save Video 4K.

Save Video 4K

images takes an IMAGE batch, a LATENT, or a VIDEO, such as Load Video 4K's video output or ComfyUI's own Load Video. A VIDEO brings its own frame rate and audio; a connected audio input still wins.

h264, hevc and av1 on NVENC, h264/hevc on CPU, prores, or hevc 10-bit (nvenc). quality is the cq/crf value, lower being better. ProRes and 10-bit HEVC are fed 16 bits per channel, so gradients keep their full 10 bits, and ProRes carries uncompressed 24-bit audio. Audio is muxed when connected, trimmed to the picture so no frames are lost. 23.976, 29.97 and 59.94 fps are written as the exact NTSC rates. Odd dimensions are padded rather than refused. The finished video plays on the node itself.

save_metadata (on by default) stores the workflow in the video file, so the result can be dragged back into ComfyUI to rebuild the graph.

vae accepts a LATENT on images and decodes it in batches, so a long 4K clip does not have to fit in VRAM in one go.

Colour

For 8-bit 4:2:0 sources the YUV to RGB conversion runs on the GPU from nv12, which moves a third of the bytes rgb24 would and is measurably closer to the source than ffmpeg's own converter, which carries a systematic -1.24/255 darkening.

Untagged files get the matrix every player assumes (bt709 at 720p and above, bt601 below) rather than ffmpeg's blanket bt601, which tints HD footage. Sources with alpha, more than 8 bits, or other chroma layouts use a matching ffmpeg conversion.

Saved files are converted with the bt709 matrix and tagged bt709 / sRGB, as ComfyUI's own Save Video does, so players show the colours that were rendered.

Requirements

ffmpeg on PATH, or imageio-ffmpeg installed. NVENC needs an NVIDIA GPU; without CUDA everything still runs on the CPU, just slower.