RelayVSRConditioning - ComfyUI Built-in Node Documentation

RelayVSR Conditioning prepares the conditioning and the empty latent for a RelayVSR video upscale. RelayVSR only runs its one-step Wan 2.2 model on every 15th frame of the clip, plus the last one, so the latent it hands you holds just those keyframes. VSR Decode then fills in the frames between them.

Inputs

Parameter	Description	Data Type	Required	Range
`model`	The RelayVSR diffusion model, loaded with Load Model: `relayvsr_x2_lightx2v_int8_convrot.safetensors` or `relayvsr_x4_int8_convrot.safetensors`.	MODEL	Yes	N/A
`images`	The low-quality video at its native resolution. The model’s upscale factor decides how large the result gets.	IMAGE	Yes	N/A

Outputs

Output Name	Description	Data Type
`positive`	Conditioning that carries the selected keyframes as low-quality guidance.	CONDITIONING
`latent`	Empty latent with one frame per keyframe, ready for the sampler.	LATENT

Usage

Sample the latent with one step at CFG 1: there is no prompt and no text encoder in the graph. Send the sampler’s output, the same images and a RelayVSR Flash VAE into VSR Decode, and keep the pair matched: the x2 model with the x2 decoder, the x4 model with the x4 decoder. Only keyframes are sampled, so a 150-frame clip turns into an 11-frame latent. The sampler works on a small fraction of the video, and VSR Decode restores the rest.

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): fd1a86262240430e4f2a811c2c9d8d62f0b6cdfccb26d5285e36d0d6de3b186f

​Inputs

​Outputs

​Usage

Inputs

Outputs

Usage