SDXL tile decoders (Core ML)

Compiled Core ML VAE decoders for a fixed 56 x 56 latent tile (448 px output). They decode an SDXL picture in overlapping tiles on devices that cannot decode the whole picture at once.

Folder VAE weights from
dreamshaper-xl-turbo/VAEDecoderTile.mlmodelc DreamShaper XL Turbo by Lykon
realvis-xl-v5/VAEDecoderTile.mlmodelc RealVisXL V5.0 by SG161222

FLOAT32 weights, about 198 MB each. Decoder only: no prompt, no encoder, nothing that makes a picture on its own.

These files are a conversion of the VAE decoders of the models above. The use restrictions of the CreativeML Open RAIL++-M License apply, see LICENSE (Attachment A) and NOTICE.md.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support