Ai art wiki is a free online encyclopedia, created and edited by volunteers. If you see wrong or missing information, feel free to edit it. Guidelines

List of text-to-image models

From AI art wiki
Jump to navigation Jump to search

General purpose[edit | edit source]

List of general purpose models
Name Description NSFW License Samples
Stable Diffusion This is a model that can be used to generate and modify images based on text prompts. It is a Latent Diffusion Model that uses a fixed, pretrained text encoder (OpenCLIP-ViT/H as of 2.1). openrail++ Merged-0005.jpg
DALL-E DALL-E is a proprietary text-to-image model. It's notable for having outpainting and filters that explicitly prevent generation of offensive content.

no Proprietary DALL-E samples.jpg
MidJourney Midjourney is a proprietary text-to-image model. One of its most prominent features is ability to combine multiple objects into one.

Proprietary Midjourney v4 hero 2.jpg
HassanBlend Photorealistic model with emphasis on drawing people with correct anatomy. Explicitly NSFW. Good for generating old people and mixing with other models.

yes creativeml-openrail-m HassanBlend.jpg

Targeted models[edit | edit source]

Targeted models are trained to draw a specific thing: a certain object or to mimic a certain artist's style.

Style[edit | edit source]

Anime[edit | edit source]

See also: Comparison of anime text-to-image models

Key Description License Extra features NSFW Samples
Waifu Waifu Diffusion is a latent text-to-image diffusion model that has been conditioned on high-quality anime images through fine-tuning.

Based on Stable Diffusion

Code: AGPL-3.0 license

Weights: CreativeML Open RAIL-M

Yes Normal.jpg
NovelAI Driven by AI, painlessly construct unique stories, thrilling tales, seductive romances, or just fool around. Anything goes!

Proprietary AI storyteller Yes Nai.jpg
Anything_v3 Welcome to Anything V3 - a latent diffusion model for weebs. This model is intended to produce high-quality, highly detailed anime style with just a few prompts. Like other anime-style Stable Diffusion models, it also supports danbooru tags to generate images.

creativeml-openrail-m Yes 1girl.png
Trinart2 Old trinart

creativeml-openrail-m No Trinart samples.jpg
Derrida Derrida (formerly TrinArt Characters v2) is a stable diffusion anime models that aims to produce pictures of characters with anatomical accuracy, but without explicit content.

creativeml-openrail-m No Derrida.jpg

Other artistic styles[edit | edit source]

Name Filename Description Nsfw Prompt Samples
wikiart-v2 sd-wikiart-v2 is a stable diffusion model that has been fine-tuned on the wikiart dataset to generate artistic images in different style and genres. The model uses tags from wikiart website. To get right prompt, go to right painting and copy what you need.

(See description) Wikiart.png
Abstract Swirls Generates pretty Junji Ito-ish swirly patterns. Its real power reveals itself when mixed with other models or is used for inpainting.

abstractswirls, long shot, by rutkowski and mucha Abstractswirls.jpg
Beksinsky the style of the mysterious and amazing artist Zdzisław Beksinski

beksinski style Beksinski.jpg
Van_Gogh The model is trained on frames from the film "Van Gogh. With love, Vincent" — which is made in the style of Van Gogh's drawing, this is the world's first animated feature film, completely painted with oil paints on boards! If you don't like too bluish gamma and yellow faces — in the negative prompt, you need to write something like yellow face, blue. The model is incredible, it draws both people and nature and buildings very well. Since the model is highly artistic, do not raise the CFG Scale high so that the neural network can "fantasize" more, 4-7 should be fine.

no lvngvncnt Vangogh1.jpgVangogh2.jpg
DBWayneBarlowe_JM Model trained on art by Wayne Barlowe. Use religious, demonic oriented prompts.

DBWayneBarlowe Wayne Barlowe JB.jpg
BubblyDubbly no art by bubblydubbly Bubblydubbly.jpg
Redshift Model trained on 3D art

no (redshift style) Redshift-diffusion-samples-02s.jpgRedshift-diffusion-samples-01s.jpg
mo-di-diffusion moDi-v1-pruned.ckpt This is the fine-tuned Stable Diffusion model trained on screenshots from a popular animation studio.

no modern disney Modi-samples-01s.jpgModi-samples-02s.jpgModi-samples-03s.jpg
classic-anim-diffusion classicAnim-v1.ckpt This is the fine-tuned Stable Diffusion model trained on screenshots from a popular animation studio.

no classic disney style Clanim-samples-01s.jpgClanim-samples-02s.jpgClanim-samples-03s.jpg
tron-legacy-style This is a fine-tuned Stable Diffusion model (based on v1.5) trained on screenshots from the film Tron: Legacy (2010).

no trnlgcy Trnlgcy-preview.jpgTrnlgcy-preview2.jpg
elden-ring-diffusion elden ring style
spider-verse-diffusion spiderverse style
disco-elysium discoelysium style
Arcane-Diffusion arcane style
Bloodborne-Diffusion This is a Dreamboothed Stable Diffusion model trained on the Bloodborne series Style.

The total dataset is made of 100 pictures, and the training has been done on runawayml 1.5 and the new VAE, with 12k steps (poly LR1e-6).

Bloodborne Style Bloodbornestyle showcase.jpg
DarkSoulsDiffusion This is a Dreamboothed Stable Diffusion model trained on the DarkSouls series Style.

The total dataset is made of 100 pictures, and the training has been done on runawayml 1.5 and the new VAE, with 2500 steps (LR1e-6) then 24k more steps (LR1e-7).

The recommended sampling is k_Euler_a or DPM++ 2M Karras on 20 steps, CFGS 7 .

DarkSouls Style
Borderlands borderlands 101670085297kl2f7g0986xphcajqydsmkvinn2dspgvecoq98y91oakvxvvpw2txcldp1kzcewxh6xni4rid5cihq44ypmbzxkxmhmkpemumfdp.png
Beeple Model generates images in Beeple style.

beeple style Beeple.png
sd-db-block-world-rtx This is the fine-tuned Stable Diffusion model trained on screenshots of Minecraft running in RTX.

BlockWorldRTX Blockworldrtx.jpg
Lozhkin Model trained on psychadelic art of Vasya Lozhkin

by Lozhkin Lozhkin sample.jpg
Synthwave Stable Diffusion model to create images in Synthwave/outrun style, trained using DreamBooth

snthwve style Synthwave.jpg
Inkpunk Finetuned Stable Diffusion model trained on dreambooth. Vaguely inspired by Gorillaz, FLCL, and Yoji Shinkawa.

nvinkpunk I3.jpg

Object[edit | edit source]

Name Filename Description Nsfw Prompt
Endless characters various Endless characters is actually a few dozens of models dedicated to portray a member of certain fantasy race. For words of creator:

For years I've hosted websites where people can go and find a suitable avatar/portrait for their roleplaying character, rather that be for Dungeons & Dragons, Ars Magica, Vampire the Masquerade, etc. In the past I've compiled other peoples art and converted them to tokens. With Stable Diffusion this allows me to create portraits and character art with the click of a few buttons. Experimenting with the prompting to see what I can get. I've created a few checkpoints for usage with the Automatic1111 GUI Stable Diffusion build, or any other GUI that allows custom checkpoints. I'll be adding more as I create them.

Prompts and examples

Gallery with thousands of examples

Close-up Portrait of _____________, Fantasy, Medieval, Volumetric lighting, concept art, brush stroke style, artstation, trending, highly detailed Endlesscharacters.jpg
James Webb Space Telescope This is a fine-tuned Stable Diffusion model (based on v1.5) trained on images taken by the James Webb Space Telescope, as well as Judy Schmidt. Use the token JWST in your prompts to use the style (e.g., "jwst, green spiral galaxy").

jwst PreviewJWST.jpg
Microscopic model V1 This is the fine-tuned Stable Diffusion model trained on microscopic images.

microscopic 1667934752243-635749860725c2f190a76e88.png
StarSector-Portraits starsectorportrait StarSector-Portraits.png
EpicSpaceMachine Generates space ships, mechanisms and mean photos of tangled paperclips and wires.

EpicSpaceMachine Photo of a futuristic space liner, 4K, award winning in EpicSpaceMachine style.jpgA photo of a tangle of wires, close up, 8K, in EpicSpaceMachine style.jpg

Model hosting sites[edit | edit source]