Method encode

nodes.py:2258–2292

(self, vae, image, enable_vae_tiling, tile_x, tile_y, tile_stride_x, tile_stride_y, noise_aug_strength=0.0, latent_strength=1.0, mask=None)

    CATEGORY = "WanVideoWrapper"

    def encode(self, vae, image, enable_vae_tiling, tile_x, tile_y, tile_stride_x, tile_stride_y, noise_aug_strength=0.0, latent_strength=1.0, mask=None):
        vae.to(device)

        image = image.clone()

        B, H, W, C = image.shape
        if W % 16 != 0 or H % 16 != 0:
            new_height = (H // 16) * 16
            new_width = (W // 16) * 16
            log.warning(f"Image size {W}x{H} is not divisible by 16, resizing to {new_width}x{new_height}")
            image = common_upscale(image.movedim(-1, 1), new_width, new_height, "lanczos", "disabled").movedim(1, -1)

        if image.shape[-1] == 4:
            image = image[..., :3]
        image = image.to(vae.dtype).to(device).unsqueeze(0).permute(0, 4, 1, 2, 3) # B, C, T, H, W        

        if noise_aug_strength > 0.0:
            image = add_noise_to_reference_video(image, ratio=noise_aug_strength)

        if isinstance(vae, TAEHV):
            latents = vae.encode_video(image.permute(0, 2, 1, 3, 4), parallel=False)# B, T, C, H, W
            latents = latents.permute(0, 2, 1, 3, 4)
        else:
            latents = vae.encode(image * 2.0 - 1.0, device=device, tiled=enable_vae_tiling, tile_size=(tile_x//vae.upsampling_factor, tile_y//vae.upsampling_factor), tile_stride=(tile_stride_x//vae.upsampling_factor, tile_stride_y//vae.upsampling_factor))

            vae.to(offload_device)
        if latent_strength != 1.0:
            latents *= latent_strength

        latents = latents.cpu()

        log.info(f"WanVideoEncode: Encoded latents shape {latents.shape}")
        mm.soft_empty_cache()

        return ({"samples": latents, "noise_mask": mask},)

NODE_CLASS_MAPPINGS = {
    "WanVideoDecode": WanVideoDecode,