Models Overview

Explore the different models available. Choose the best model for your specific image generation needs.

Introduction

A model is a pre-trained neural network that generates images based on the data it has been trained on. Choosing the right model is crucial as it directly impacts the quality and style of the generated images.

Different models provides different advantages. Models allow you to experiment with different artistic styles and leverage specialized capabilities such as ultra-realistic rendering or imaginative and dreamy creations. This flexibility ensures that you can consistently meet your application's specific goals and deliver compelling visual content.

LoRAs and ControlNet are also supported in our platform. These specialized models allow you to enhance the output of the generated images, providing styling and control to tailor the images to your specific needs.

Mix and match multiple LoRAs and ControlNet models to create unique and engaging images.

AIR system

At Runware we don't offer a limited set of curated models. Instead, we provide the ability to use any model available via the AIR (Artificial Intelligence Resource) system. This system allows you to access a wide range of models from different providers, ensuring that you can always find the best model for your specific needs.

This list is constantly growing, so you can find the latest models available in the Model Explorer that you will find below.

CivitAI is one of the most popular websites that uses AIR identifiers, so any model you find there can be used in our platform, just copy the AIR identifier and use it in your requests.

Model metadata from CivitAI — On CivitAI website, you can find the AIR identifier in the right sidebar of the model page

If by any chance you can't find the model you are looking for, you can always upload your own model to our platform and tag it with the AIR system. This way you can use it in the same way as any other model available.

Model Explorer

We created an interactive Model Explorer so you can search for the best model for your specific needs and get the model identifier to use in your API requests.

checkpoint SD1.5

epiCRealism Natural Sin RC1 VAE
photorealistic, female, analog, realism, base model, portrait, raw photo, woman, photography, photorealism, realistic, photoshoot, natural colors
checkpoint SD1.5

GhostMix v2.0-BakedVAE
anime, photorealistic, female, base model, art style, game character, girls, portraits, realistic
lora SD1.5 inpainting

GHIBLI_Background END
ghibli, scenery, background
checkpoint SD1.5

ToonYou Beta 5 🌟 Unstable
anime, base model, comics, 2d, cartoon, girls, pretty, comic
checkpoint SD1.5

DreamShaper v1
anime, landscapes, 3d, photorealistic, inpainting, digital art, scifi, base model, fantasy art, art style, paintings, woman, illustration, 2d, digital illustration, fantasy, girls, portraits, realistic, video game
lora SD1.5 inpainting

Ink scenery | 水墨山水 v1.0
water, ink, scenery, background, tree, mountain
checkpoint SD1.5

ReV Animated V2 Rebirth
anime, base model, illustration, cartoon, fantasy, portraits
checkpoint sdxl

Juggernaut XL Jugg_X_RunDiffusion_Hyper
photorealistic, base model, photo, cinematic, characters, nsfw
lora SD1.5 inpainting

墨心 MoXin 墨心 MoXin 1.0
chinese, chinese dress, style, hanfu, paintings, artstyle
checkpoint sdxl

RealVisXL V4.0 V4.0 (BakedVAE)
photorealistic, base model, woman, man, realistic, semi-realistic, nsfw, sfw
lora sdxl inpainting

JJ's Interior Space - Bathroom XL v1.0
buildings
checkpoint SD1.5

AbsoluteReality v1.8.1
landscapes, photorealistic, base model, portrait, hyperrealistic character portraits, photography, realistic
lora SD1.5 inpainting

KIDS ILLUSTRATION COOLKIDS
simple, style, illustration, cartoon, books, kids
checkpoint sdxl

SocaRealism XL 4.0
base model, woman, realistic, general
checkpoint SD1.5

dvArch - Multi-Prompt Architecture Tuned Model dvArch: Exterior
architecture, victorian, buildings, gothic, modern, interior design, exterior

Basic ControlNet models

We also provide a curated list of basic/common ControlNet models that you can use without the need to search for them in the Model Explorer.

You can use this models directly by specifying their name in the model parameter inside a controlNet object.

Name	SD 1.5	SDXL	Description
`canny`	✅	✅	Applies the Canny edge detection algorithm.
`depth_leres`	✅	✅	Generates depth maps using the Leres method.
`depth_midas`	✅	✅	Uses the MiDaS model to estimate depth from a single image.
`depth_zoe`	✅	✅	Depth estimation using the Zoe model.
`inpaint_global_harmonious`	✅	✅	Harmonious inpainting technique for global image completion.
`lineart_anime`	✅		Converts images into anime-style line art.
`lineart_coarse`	✅		Produces coarse line art from images.
`lineart_realistic`	✅		Generates realistic line art from images.
`lineart_standard`	✅		Standard line art generation from images.
`mlsd`	✅		Generates multi-level semantic segmentation masks.
`normal_bae`	✅		Estimates surface normals using the BAE method.
`openpose`	✅		Detects human keypoints and poses in images using the OpenPose model.
`openpose_face`	✅		Detects poses + facial keypoints.
`openpose_faceonly`	✅		Detects facial keypoints only.
`openpose_full`	✅		Detects poses + facial keypoints + hand keypoints.
`openpose_hand`	✅		Detects hand keypoints.
`scribble_hed`	✅		Produces edge maps from user-provided scribbles using the HED model.
`scribble_pidinet`	✅		Edge map generation using the PIDinet model from user-provided scribbles.
`seg_ofade20k`	✅		Semantic segmentation using the OFA dataset with 20k categories.
`seg_ofcoco`	✅		Semantic segmentation using the COCO dataset.
`seg_ufade20k`	✅		Semantic segmentation using the UFA dataset with 20k categories.
`shuffle`	✅		Randomly shuffles images.
`softedge_hed`	✅		Soft edge detection using the HED model.
`softedge_hedsafe`	✅		Safe variant of soft edge detection using the HED model.
`softedge_pidinet`	✅		Soft edge detection using the PIDinet model.
`softedge_pidisafe`	✅		Safe variant of soft edge detection using the PIDinet model.
`tile_gaussian`	✅		Applies Gaussian tiling to images.