ReCaVSR

ReCaVSR is a one-step streaming diffusion model for video super-resolution. See the GitHub repository for inference instructions.

License

The files listed below are distributed under the Apache License 2.0. See LICENSE for the full text.

File	Provenance
transformer.safetensors	Modified from Wan2.2-TI2V-5B, with ReCaVSR's LoRA weights merged and learned LR/SR conditioning projections added.
flashdecoder.safetensors	ReCaVSR's separately trained, low-resolution-conditioned decoder, based on the FlashDecoder architecture.
prompt.safetensors	Precomputed prompt embedding used for ReCaVSR inference.
vae.safetensors	Unchanged copy of the official Wan2.2 VAE checkpoint, distributed under Wan2.2's upstream Apache-2.0 license.