Clarify VAE use

README.md

@@ -9,7 +9,9 @@ tags:
 
 This model was generated by Hugging Face using [Apple’s repository](https://github.com/apple/ml-stable-diffusion) which has [ASCL](https://github.com/apple/ml-stable-diffusion/blob/main/LICENSE.md). This version contains Core ML weights with the `ORIGINAL` attention implementation, suitable for running on macOS GPUs.
 
-This version uses 4.04 mixed-bit palettization and generates images with a resolution of 768×768. It uses `SPLIT_EINSUM` attention and is intended for use in iOS/iPadOS 17 or better. For a version better suited for macOS, please check [this model](https://huggingface.co/apple/coreml-stable-diffusion-xl-base) or [the mixed-bit palettization resources](https://huggingface.co/apple/coreml-stable-diffusion-mixed-bit-palettization).
+This version uses 4.04 mixed-bit palettization and generates images with a resolution of 768×768. It uses `SPLIT_EINSUM` attention and is intended for use in iOS/iPadOS 17 or better. It also uses a version of the VAE decoder [created by `@madebyollin`](https://huggingface.co/madebyollin/sdxl-vae-fp16-fix), which allows running using 16-bit precision and helps keep the memory footprint contained.
+
+For a macOS version, please check [this model](https://huggingface.co/apple/coreml-stable-diffusion-xl-base) or [the mixed-bit palettization resources](https://huggingface.co/apple/coreml-stable-diffusion-mixed-bit-palettization).
 
 Check [the original repository](https://github.com/apple/ml-stable-diffusion) for benchmark numbers on various devices.