CyberRealistic V9 (Core ML 8-bit)

Cyberdelia's CyberRealistic V9 as an 8-bit Core ML bundle for making photorealistic pictures on iPhone and iPad with TokForge.

Published by TokForge for on-device inference.

What it is Image generator (photorealistic)
Format Core ML resources for Apple's ml-stable-diffusion in one zip, 8-bit palettized, split einsum attention, 512 x 512
Runs on iPhone and iPad (Core ML, Apple Neural Engine); listed in TokForge for iOS
Phone class 4 GB RAM or more, iOS 17 or later (TokForge for iOS catalog)
Download 1.11 GB (1 file)
Base model cyberdelia/CyberRealistic
License CreativeML OpenRAIL-M, plus Cyberdelia's Civitai terms

TokForge

Runs on-device in the TokForge app.

Use it in TokForge

  1. Join the iOS TestFlight to install TokForge on iPhone or iPad.
  2. TokForge for iOS lists it in its image models as CyberRealistic V9 ยท 8-bit (ANE). The app marks it as a mature (photoreal) model, so it shows up only after mature models are turned on. Pick it and the app downloads the zip, unpacks it and runs it on the Neural Engine.
  3. Describe the picture you want and generate it. The model runs on the device; once it is downloaded, making pictures needs no connection.

Collections: TokForge iOS Models.

Files

File Size What it is
CyberRealistic-V9_palettized-8bit_split_einsum_compiled.zip 1.11 GB Compiled Core ML resources (a Resources/ folder) in one zip; the file TokForge downloads

The zip holds a Resources/ folder with TextEncoder.mlmodelc, Unet.mlmodelc, VAEDecoder.mlmodelc, VAEEncoder.mlmodelc and the CLIP tokenizer files (vocab.json and merges.txt). Unzip it to get the layout Apple's StableDiffusionPipeline and the TokForge installer load. Unpacked, it takes about 1.19 GB.

About the model

CyberRealistic V9 is cyberdelia's photorealistic Stable Diffusion 1.5 fine-tune with an integrated VAE. This bundle is made from its CyberRealistic_V9_FP16 checkpoint.

This is the higher-precision sibling of TokForge-CyberRealistic-V9-CoreML-6bit: a larger download, closer to the FP16 original. It is meant for prompts where the 6-bit build shows rainbow or banding artifacts.

Architecture Stable Diffusion 1.5 (UNet, CLIP text encoder, VAE)
Weights UNet and text encoder 8-bit palettized, VAE in fp16
Components TextEncoder, Unet, VAEDecoder and VAEEncoder (.mlmodelc), plus the CLIP tokenizer files
Output 512 x 512, fixed in the compiled model

Render settings

Setting Value
Attention split einsum (Apple Neural Engine)
Compute units .cpuAndNeuralEngine
Steps 25 to 30 (TokForge starts at 28)
Guidance (CFG) 7.0
Resolution 512 x 512 (SD 1.5 native; baked into the compiled model)

How it was made

Converted from the CyberRealistic_V9_FP16 checkpoint with Apple ml-stable-diffusion (torch2coreml), with split einsum attention for the Neural Engine and 8-bit palettization (--quantize-nbits 8), then bundled as compiled resources for the Swift pipeline. TokForge uses it on iOS 17 or later.

Limitations

  • Pictures are 512 x 512 only; the size is built into the compiled model.
  • The download is larger than the 6-bit build (1.11 GB against 917 MB).
  • This is a Core ML runtime bundle, not a diffusers or PyTorch checkpoint.

License

CyberRealistic V9 and this conversion are under the CreativeML OpenRAIL-M license, inherited from CyberRealistic / Stable Diffusion 1.5. Use is subject to the OpenRAIL-M restrictions. Cyberdelia's Civitai terms apply too (see Credits).

This repository is a redistribution for on-device use, a format conversion (PyTorch -> Core ML) and 8-bit palettization of cyberdelia's CyberRealistic V9. No weights were retrained. The original OpenRAIL-M terms and attribution requirements carry over to this conversion and to any images generated with it. This repackaging adds no further restrictions.

Credits

  • CyberRealistic V9 by Cyberdelia (Civitai, Hugging Face). Used under the CreativeML OpenRAIL-M license and the creator's Civitai terms. Cyberdelia's Civitai terms allow selling and do not add a credit requirement; the CreativeML OpenRAIL-M terms still apply.
  • Stable Diffusion v1.5 by Runway and CompVis (Hugging Face). Used under the CreativeML OpenRAIL-M license.
  • Converted to Core ML with ml-stable-diffusion by Apple (MIT).
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

Model tree for darkmaniac7/TokForge-CyberRealistic-V9-CoreML-8bit

Quantized
(2)
this model

Collection including darkmaniac7/TokForge-CyberRealistic-V9-CoreML-8bit