SDXL

What is SDXL Uncensored?

SDXL (Stable Diffusion XL) is an open image model released by Stability AI in July 2023. "SDXL Uncensored" usually means SDXL run without a content filter — on your own GPU or on a service that hosts it — often with community fine-tunes and LoRAs for adult content.

The base model itself is open-weight, so anyone can download it, run it locally and change it. That openness is what made it popular: thousands of community LoRAs were trained on top of it for styles, characters and body types.

Is SDXL available in ZenCreator?

No. SDXL is not in ZenCreator's model menus anymore — not in Text-to-Image, Reference-to-Image, the Image Editor or Head & Face Swap.

Its place was taken by newer models:

• Krea 2 replaced SDXL in Reference-to-Image and is also in Text-to-Image.
• Seedream 5 Pro is the other Reference-to-Image model, and is also in Text-to-Image and the Image Editor.

The cards at the top of this page link straight to each tool.

How does SDXL work?

SDXL generates an image in two stages. A base model creates the image, then a separate refiner model cleans up the fine detail.

• Size: 3.5 billion parameters in the base model, 6.6 billion with the refiner.
• Resolution: native 1024×1024 — four times the pixel area of Stable Diffusion 1.5 (512×512).
• Prompt reading: two text encoders (OpenCLIP ViT-bigG and CLIP ViT-L) read the prompt together, so it follows prompts better than earlier Stable Diffusion versions.
• Hardware: it runs on consumer GPUs with about 8 GB of VRAM.

Why did creators pick SDXL?

Three reasons, in order of how often they come up:

1. LoRAs. SDXL has one of the largest libraries of community LoRAs — small add-on files that teach the model a style, a character or a body type without retraining it.
2. Cost. It is small enough to run cheaply, so it suited long runs of test generations.
3. Control. Open weights mean no provider-side filter when you host it yourself.

The trade-offs are real too. SDXL is a 2023 design: it handles complex scenes with several people or exact layouts worse than newer models, and stacking strong LoRAs often breaks anatomy — extra fingers, bent limbs.

Which ZenCreator model should you use instead of SDXL?

Pick by what you used SDXL for.

All four models are in the current model menus. Credits are the same ones you use for every tool, and they never expire.

Questions

Can I still use SDXL in ZenCreator?

No. SDXL is no longer in any ZenCreator tool, including the deprecated-models list.

For reference-based images, Krea 2 is the direct replacement in Reference-to-Image. For text prompts, use Krea 2 or Seedream 5 Pro in Text-to-Image.

Do my SDXL LoRAs work with Krea 2 or Flux Klein?

No. A LoRA is trained for one model architecture, so an SDXL LoRA does not load into Krea 2 or Flux Klein.

In ZenCreator, LoRA styles now come as ready presets in Flux Klein + LoRA in the Image Editor. Krea 2 uses its own built-in style LoRA.

Is SDXL still worth running locally?

Yes, if you rely on a specific SDXL LoRA or fine-tune. It runs on GPUs with about 8 GB of VRAM, and its community library is still large.

If you do not need a particular LoRA, newer models give cleaner anatomy and follow complex prompts better without any setup.

Which replacement is closest to SDXL's photo-only reference mode?

Krea 2 in Reference-to-Image. Like SDXL there, it takes just one photo and returns a similar image.

The prompt is optional: leave it empty and the reference is described automatically. Use the strength slider to choose how much of the photo is redrawn.

Sources

1. Stability AI — SDXL 1.0 release announcement: stability.ai
2. Podell et al., "SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis" (2023): arxiv.org/abs/2307.01952
3. ZenCreator app model menus and public API tool list, checked October 2026