LocalForge AILocalForge AI
LibraryBlogFAQ

Civitai Models Download List - Free, Uncensored, Runs Locally

Every model on this list is free, uncensored, and runs 100% on your hardware. No subscriptions, no cloud, no content filters. We picked 10 models across Flux, SDXL, and SD 1.5 - each with a direct CivitAI link, file size, and minimum VRAM requirement.

Before you click download: skim license/usage notes on each Civitai card (commercial vs non-commercial varies). After files land on disk, use how to download Civitai models for folder hygiene, then how to run Civitai models locally or use Civitai models in ComfyUI so the first generation actually runs.

Downloads

1. Fluxed Up

Top Pick

Most downloaded Flux NSFW model. Best absolute quality if you have the VRAM.

Architecture: Flux · VRAM: 12 GB+ · Best for: Photorealistic NSFW

Download →

2. Juggernaut XL v9

The default pick for photorealism. 1.4M downloads.

Architecture: SDXL · VRAM: 8 GB+ · Best for: All-round photorealism

Download →

3. RealVisXL V5.0

Sharpest facial detail of any SDXL model.

Architecture: SDXL · VRAM: 8 GB+ · Best for: Portraits

Download →

4. Persephone

Handles both SFW and NSFW equally well.

Architecture: Flux · VRAM: 12 GB+ · Best for: NSFW/SFW hybrid

Download →

5. epiCRealism XL

Best texture detail for close-up shots.

Architecture: SDXL · VRAM: 8 GB+ · Best for: Beauty/editorial

Download →

6. CyberRealistic v9

Only good realistic model that runs on 4 GB VRAM.

Architecture: SD 1.5 · VRAM: 4 GB+ · Best for: Low-VRAM realism

Download →

7. Pony Diffusion V6 XL

Largest LoRA ecosystem. Most versatile anime model.

Architecture: SDXL · VRAM: 8 GB+ · Best for: Anime/character NSFW

Download →

8. DreamShaper XL

Best for blending realistic and artistic styles.

Architecture: SDXL · VRAM: 8 GB+ · Best for: Fantasy/sci-fi

Download →

9. WAI-illustrious-SDXL

Best anime model. 19K+ ratings, active updates.

Architecture: Illustrious · VRAM: 8 GB+ · Best for: Anime illustration

Download →

10. Z-Image Turbo

LoRA that speeds up Flux NSFW generation.

Architecture: Flux · VRAM: 12 GB+ · Best for: Fast Flux NSFW

Download →

Quick Reference

Model Arch Size VRAM Best For Download
Fluxed Up Flux ~12 GB 12 GB+ Photorealistic NSFW CivitAI
Juggernaut XL v9 SDXL ~6.5 GB 8 GB+ All-round photorealism CivitAI
RealVisXL V5.0 SDXL ~6.5 GB 8 GB+ Portraits CivitAI
Persephone Flux ~12 GB 12 GB+ NSFW/SFW hybrid CivitAI
epiCRealism XL SDXL ~6.5 GB 8 GB+ Beauty/editorial CivitAI
CyberRealistic v9 SD 1.5 ~2 GB 4 GB+ Low-VRAM realism CivitAI
Pony Diffusion V6 XL SDXL ~6.5 GB 8 GB+ Anime/character NSFW CivitAI
DreamShaper XL SDXL ~6.5 GB 8 GB+ Fantasy/sci-fi CivitAI
WAI-illustrious-SDXL Illustrious ~6.5 GB 8 GB+ Anime illustration CivitAI
Z-Image Turbo Flux varies 12 GB+ Fast Flux NSFW CivitAI

Checkpoints - Photorealistic

Fluxed Up (Flux NSFW Checkpoint)

Most downloaded Flux NSFW model - 8.6 million downloads.

If you have 12+ GB VRAM, this produces the most realistic images of any model on CivitAI. The tradeoff: Flux models need more GPU memory than SDXL, and this one uses a non-commercial license.

Juggernaut XL v9

Best all-round SDXL model - 1.4 million downloads, 3,025 reviews.

The default pick for photorealism. Cinematic lighting, smooth skin, good hands. The Ragnarok update (May 2025) fixed most hand/feet issues. Still struggles with text rendering.

RealVisXL V5.0

Best for portraits - 2 million+ downloads.

Sharper facial detail than Juggernaut, especially for close-up shots. Includes Lightning variants for faster generation. Less versatile for non-portrait scenes.

Persephone (Flux NSFW/SFW)

Versatile Flux hybrid - works clothed and unclothed.

Handles both SFW and NSFW equally well, which is rare for Flux models. 1.1 million downloads. Non-commercial license.

epiCRealism XL

Best texture detail - editorial and beauty photography.

Skin texture, pores, hair detail - epiCRealism beats everything else in close-up shots. Less flexible than Juggernaut for full-body or environmental scenes.

CyberRealistic v9

Best option if you only have 4 GB VRAM.

SD 1.5 is an older architecture, but CyberRealistic squeezes impressive results from it. Regularly updated (last: March 2026). The only realistic model that runs on budget GPUs like the GTX 1650.

Checkpoints - Anime / Character

Pony Diffusion V6 XL

Largest LoRA ecosystem - 2.6 million training images, fully uncensored.

The most versatile anime model. Supports anime, cartoon, furry, and more. Requires specific prompt tags (score_9, score_8_up) for best results. Massive community with thousands of compatible LoRAs.

DreamShaper XL

Bridges realistic and artistic - fantasy, sci-fi, concept art.

Pick DreamShaper when you want something between photorealism and illustration. Great for fantasy characters, sci-fi environments, and concept art.

WAI-illustrious-SDXL

Top anime model - 19,243+ ratings.

Currently the best anime illustration model on CivitAI. Active development (v16 latest), huge community, best character accuracy of any anime model. Uses quality score tags like Pony.

LoRAs

Z-Image Turbo (Flux NSFW LoRA)

Fast Flux NSFW with fewer steps.

Reduces Flux generation time by cutting step count. Apply this on top of Fluxed Up or another Flux checkpoint for faster NSFW output.

How to Install These Models

All files are .safetensors format. Download and place in the correct folder:

  • Forge / A1111: models/Stable-diffusion/ for checkpoints, models/Lora/ for LoRAs
  • ComfyUI: models/checkpoints/ for checkpoints, models/loras/ for LoRAs

Then refresh your model list in the UI.

Need the full walkthrough? See How to Download Civitai Models for step-by-step instructions, or How to Run Civitai Models Locally for complete setup from scratch.

Or use LocalForge AI for Forge pre-configured with popular models and zero setup.

What to Do Next

What to Do Next

First time?

Account setup, file types, folder placement

Need a UI?

Forge/ComfyUI install + first generation

Want realistic models?

Deep dive on Juggernaut, RealVisXL, and more

Want anime models?

Pony, Illustrious, Animagine compared

FAQ

Are all these models free to download? +
Yes. Every model on this list is free on CivitAI. Flux-based models use a non-commercial license, but downloading and personal use costs nothing.
What's the difference between .safetensors and .ckpt files? +
.safetensors is the modern format - faster to load and can't contain hidden code. Always pick .safetensors when you have the choice. .ckpt is the older format and carries a small security risk.
Where do I put downloaded model files? +
In Forge or A1111, checkpoints go in models/Stable-diffusion/ and LoRAs in models/Lora/. In ComfyUI, checkpoints go in models/checkpoints/ and LoRAs in models/loras/. Then refresh the model list in your UI.
How much VRAM do I need? +
SD 1.5 models (like CyberRealistic) run on 4 GB. SDXL models need 8 GB minimum. Flux models need 12 GB+. Check the VRAM column in the table above to match your GPU.
Can I use these models commercially? +
SDXL and SD 1.5 models generally have open licenses. Flux-based models (Fluxed Up, Persephone, Z-Image Turbo) use the Flux.1 D non-commercial license. Check each model's CivitAI page for the specific license.