Flux Video Upscale - ComfyUI Built-in Node Documentation

Flux Video Upscale enlarges a video with FLUX super-resolution, either to a fixed 1080p, 2K or 4K frame size or by 1.5 to 3 times. Creative mode restores and invents fine detail, which suits generated footage; precise mode sharpens the source without changing it, which suits faces, products and real footage. The clip is uploaded, the BFL job is polled, and the finished video comes back at 24 fps.

Inputs

Common Inputs

Parameter	Description	Data Type	Required	Range
`video`	Source clip of 1 to 20 seconds and at most 50 MB, at least 64x64 pixels, with an aspect ratio between 1:4 and 4:1. The output is rendered at 24 fps and capped at about 14.4 megapixels per frame.	VIDEO	Yes	1 to 20 seconds duration; aspect ratio between 1:4 and 4:1; minimum 64x64 pixels
`target_resolution`	Output size. `"1080p"`, `"2K"` and `"4K"` fit the result inside 1920x1080, 2560x1440 or 3840x2160 pixels, turned for portrait clips; `"by factor"` uses `upscale_factor` instead.	DYNAMIC_COMBO	Yes	`"2K"` `"4K"` `"1080p"` `"by factor"`
`mode`	`"creative"` restores and invents fine detail, best for generated footage, textures and scenery. `"precise"` sharpens the source without changing it, for faces, products and real footage. (default: `"creative"`)	COMBO	Yes	`"creative"` `"precise"`
`prompt`	Optional description of the clip that steers the enhanced detail. Leave empty for a neutral upscale. (default: empty)	STRING	Yes	Multiline text
`auto_downscale`	Downscale sources that are too large instead of rejecting them: above 2560x1440 pixels in area, or above two thirds of the target resolution. Aspect ratio is preserved. (default: true)	BOOLEAN	Yes	true false
`safety_tolerance`	Moderation tolerance, 0 is the strictest. (default: 2, advanced parameter)	INT	Yes	0 to 4
`seed`	Seed controls whether the node should re-run; FLUX picks its own seed, so actual results are nondeterministic regardless of this value. (default: 42)	INT	Yes	0 to 4294967295

by factor Inputs

These inputs appear when target_resolution is set to "by factor".

Parameter	Description	Data Type	Required	Range
`upscale_factor`	Output size relative to the source. Very large sources are upscaled by less than the requested factor because of the per-frame cap. (default: 2.0)	FLOAT	Yes	1.5 to 3.0 (step 0.1)

A target size is applied as an upscale factor clamped to 1.5-3.0, so a source too small to fill the box at 3x comes out smaller than the requested resolution, and a source that would need less than 1.5x to fit the box is downscaled first. With auto_downscale switched off the node stops with an error instead of downscaling. The output is 24 fps and capped at about 14.4 megapixels per frame, so very large sources may be upscaled by less than the requested factor.

Outputs

Output Name	Description	Data Type
`video`	The upscaled video clip.	VIDEO

This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub

Source fingerprint (SHA-256): bef6ef4614d75bc9e6bab977f5f19c45fe30d300907913462a058e6e7e7e4ac7

​Inputs

​Common Inputs

​by factor Inputs

​Outputs

Inputs

Common Inputs

by factor Inputs

Outputs