Will It Run AI

RealVisXL v5.0

Stable

by SG161222

The most popular photorealistic SDXL fine-tune on CivitAI. Excels at lifelike portraits, landscapes, and product photography. Compatible with all SDXL ControlNets and LoRAs.

VRAM requirements, GPU fit, and setup notes for RealVisXL v5.0, including 8GB/12GB fit guidance where relevant. Recommended runtimes: ComfyUI, Automatic1111, and Diffusers support. Best download size: ~6.9 GB at FP16.

  • Top photorealistic SDXL fine-tune on CivitAI
  • Excellent portraits and landscapes
  • Full SDXL ControlNet and LoRA compatibility
  • Drop-in replacement for SDXL base
ComfyUI, Automatic1111, DiffusersFP16 safetensors

Your hardware

Detecting...

Parameters2.6B
Max Resolution1024×1024
Default Steps25
ArchitectureUNET
Licenseopenrail++

Image Quality Benchmarks

Measured quality metrics for RealVisXL v5.0 outputs.

Human Preference Score80%

How often humans prefer this model's output (0-100%)

Aesthetic Score7.8

Visual quality and composition rating (5-9 scale)

CLIP Score0.29

Text-image alignment accuracy (higher is better)

VRAM Requirements by Resolution and Precision

Compare which GPUs can run RealVisXL v5.0 at different precisions. FP8 uses less memory than FP16 when available, and the grade shows how comfortably each GPU handles the workload.

FP16 (full precision)

ResolutionVRAM RequiredRTX 4090 24GBRTX 3060 12GBRTX 4060 8GBMacBook Pro M4 Pro 24GB
512×5127.6 GBSSBS
768×7687.8 GBSSBS
1024×10248.0 GBSSBS

Optimization Tips

Turbo / LCM distillation

Use distilled scheduler at 4-8 steps for faster iteration

ControlNets available

Add guided generation with 3 adapters (+1.2 GB VRAM each)

Rich LoRA ecosystem

Customize style, characters, and quality with community LoRAs

Run with Python

Run with Python (diffusers)
from diffusers import StableDiffusionXLPipeline
import torch

pipe = StableDiffusionXLPipeline.from_pretrained(
    "SG161222/RealVisXL_V5.0",
    torch_dtype=torch.float16
)
pipe.to("cuda")

image = pipe(
    prompt="your prompt here",
    num_inference_steps=25,
    guidance_scale=7.5,
    height=1024,
    width=1024,
).images[0]
image.save("output.png")

Get started

Setup instructions for running RealVisXL v5.0 locally

1. Download the model

Get the checkpoint from HuggingFace

2. Place in:

ComfyUI/models/checkpoints/

3. Launch ComfyUI

python main.py
Tip: For SDXL fine-tunes, you can optionally add the SDXL refiner for improved detail. Place the refiner checkpoint in the same folder and add a second KSampler with denoise ~0.3.

ComfyUI Workflow

Basic txt2img workflow for RealVisXL v5.0

7 nodes

Drag & drop into ComfyUI or use File → Import

Memory Breakdown

VRAM allocation at 1024×1024 on RTX 4090 24GB (24 GB)

Required: 8.0 GBAvailable: 24.0 GB
Weights5.2 GB
VAE0.2 GB
Text Encoder1.6 GB
Activations0.5 GB
Overhead0.5 GB

Estimated Generation Time

Time per image at 1024×1024, 28 steps, FP16.

RTX 4090 24GB~4.5s
RTX 3060 12GB~17s
RTX 4060 8GB~1m 8s
MacBook Pro M4 Pro 24GB~36.4s

Sample Outputs

Available Formats, Downloads & Setup

Download RealVisXL v5.0 in the precision that matches your GPU. Lower precision usually means less VRAM pressure, while higher precision keeps more quality.

格式精度大小提供商
官方权重
safetensors推荐FP166.9 GBSG161222下载
社区转换
safetensors推荐社区FP166.9 GBcommunity-civitai下载

ControlNet Support

3 ControlNets available for RealVisXL v5.0. ControlNets add guided image generation (edges, depth, pose) at the cost of extra VRAM.

Canny Edge (SDXL)

+1.2 GB VRAM

Inherits SDXL base model ControlNet compatibility. Edge-based structural guidance.

comfyuiautomatic1111diffusers
View on HF

Depth Map (SDXL)

+1.2 GB VRAM

Inherits SDXL base model ControlNet compatibility. Depth-based spatial control.

comfyuiautomatic1111diffusers
View on HF

OpenPose (SDXL)

+1.2 GB VRAM

Inherits SDXL base model ControlNet compatibility. Human pose control.

comfyuiautomatic1111diffusers
View on HF

LoRA Ecosystem

Large Ecosystem

Inherits the full SDXL LoRA ecosystem. All SDXL LoRAs work with RealVisXL.

Approximately 5,000 LoRAs available on CivitAI. Each LoRA adds ~0.2 GB VRAM.

Browse all LoRAs on CivitAI
Fine-tune of sdxl-1-0 · Source: civitai

Related Workflows

You might also like

Frequently asked questions

FAQ — RealVisXL v5.0 VRAM, Runtimes & Fit

How much VRAM does RealVisXL v5.0 need?

RealVisXL v5.0 (2.6B parameters) requires approximately 8.0 GB of VRAM at FP16 precision for standard 1024×1024 image generation. If you want a lighter setup, lower precisions like FP8 can reduce memory use when available.

Can I run RealVisXL v5.0 on an 8GB GPU?

RealVisXL v5.0 usually needs more than 8GB for comfortable local use. Check the VRAM table above for the exact resolution and precision trade-off.

Does RealVisXL v5.0 work in ComfyUI and Automatic1111?

RealVisXL v5.0 is marked for ComfyUI, Automatic1111, and Diffusers support in our catalog, so those are the runtimes we recommend first for local setup. If your workflow uses another front end, check the model's available formats and workflow notes above before downloading.

Can I run RealVisXL v5.0 on RTX 4090?

Yes, the RTX 4090 (24 GB VRAM) can run RealVisXL v5.0 comfortably at FP16. Expected generation time is around ~4.5s per image at 1024×1024.

Does RealVisXL v5.0 support ControlNet?

Yes, RealVisXL v5.0 has 3 ControlNet adapters available: Canny Edge (SDXL), Depth Map (SDXL), OpenPose (SDXL). Each ControlNet adds roughly 1.2 GB of extra VRAM.

Does RealVisXL v5.0 have LoRA support?

Inherits the full SDXL LoRA ecosystem. All SDXL LoRAs work with RealVisXL. The LoRA ecosystem for RealVisXL v5.0 is rated as "large". There are approximately 5,000 LoRAs available on Civitai. Each LoRA adds roughly 0.2 GB of extra VRAM.

How fast is RealVisXL v5.0?

On a reference GPU (RTX 4090 24GB), RealVisXL v5.0 generates a 1024×1024 image in approximately ~4.5s at FP16 with 28 inference steps. Faster GPUs with higher memory bandwidth will produce images more quickly.

About RealVisXL v5.0

Use cases
photorealisticportraitlandscape
Recommended runtimes
comfyuiautomatic1111diffusers

See also