license: openrail++ base_model:
- stabilityai/stable-diffusion-xl-base-1.0 pipeline_tag: text-to-image tags:
- stable-diffusion
- stable-diffusion-xl
- sdxl
- text-to-image
- checkpoint
- checkpoint-merge
- safetensors
- fp16
- illustration
- graphic-novel
- comic-book
- european-comics
- concept-art
- character-art
- lineart
- sketch
- ink
- matte-painting
CHEYENNE v2.0 — SDXL Checkpoint
Model description
CHEYENNE v2.0 is an SDXL checkpoint merge created for illustration, European-style comics, graphic novels, concept art, character design, lineart, painting, and matte-painting workflows.
Version 2.0 introduces a more mature and painterly rendering style, more precise linework, improved visual versatility, and stronger support for production-oriented concept development.
The model works well with short, literal prompts. Style terms can be added to steer the output away from photographic rendering and toward illustration or comic-book aesthetics.
Model details
- Type: Checkpoint Merge
- Base model: Stable Diffusion XL 1.0
- Version: 2.0
- Precision: fp16
- Format: SafeTensor
- VAE: baked into the checkpoint
- Original author: Aurety
- Original source: Civitai
- Civitai AIR:
civitai:198051@1055511
Model file
_Cheyenne_2.0_VAE_Baked_.fp16.safetensors
Installation — ComfyUI
Place the checkpoint in:
ComfyUI/models/checkpoints/
Reload the model list or restart ComfyUI, then select the checkpoint in the Load Checkpoint node.
Recommended settings
Suggested starting configuration:
- Sampler: DPM++ 2M
- Steps: 10–30
- CFG: 1–5
Alternative samplers:
- DPM++ 2M SDE
- DPM++ 3M SDE
The SDE samplers may produce crisper and more polished images, although they can follow the prompt slightly less strictly.
Cheyenne is sensitive to sampler, step count, and CFG. Low CFG values generally produce more creative, painterly, and concept-oriented results.
Suggested prompt terms
No mandatory activation word is required.
Useful style terms include:
graphic novel
comic book style
European comics
comics
illustration
lineart
ink
sketch
concept art
character design
painting
matte painting
Add terms such as graphic novel illustration, comics, lineart, or ink when the result appears more photographic than intended.
VAE note
Although this checkpoint contains a baked VAE, the original author reported potential issues with that VAE and recommends using the external SDXL VAE FP16 Fix when necessary.
For ComfyUI, place the external VAE file in:
ComfyUI/models/vae/
Then select it explicitly in the workflow instead of relying on the baked VAE.
Recommended filename:
sdxl_vae.safetensors
Known limitations
The model may occasionally produce:
- imperfect hands or anatomy;
- outputs that are more photographic than expected;
- different results depending heavily on sampler and CFG;
- reduced prompt adherence with some SDE samplers.
Prompting explicitly for illustration, lineart, comics, or graphic-novel aesthetics can help maintain the intended visual style.
License
This checkpoint is distributed under the CreativeML Open RAIL++-M License, following the license indicated by the original publication and the SDXL 1.0 base model.
Use the model responsibly and comply with the license restrictions, applicable laws, platform policies, and third-party rights.
Attribution
Original model created by Aurety and originally published on Civitai.
This repository is not the original publication and does not claim authorship of the checkpoint weights.
Download
Download the .safetensors checkpoint from the Files and versions tab.