🎮 Niche treasure: StableDiffusion game scene, RPG model sharing

🎮 Niche treasure: StableDiffusion game scene, RPG model sharing

May 30th, 2023
View OriginalTranslated by Google

The third issue of SD special model sharing is here, and a relatively small topic of game models has been excavated. Although the download volume is not as high as that of real people and two-dimensional, all of them are treasure models. Whoever uses them will enjoy them!


Features : universal model, strong divergence and creativity

One of the top models on the Civitai website, the download volume has reached 0.2M. The author said that this is a model compatible with any case. The requirement is that the more detailed the prompts, the better, and can use professional vocabulary related to painting; the simpler the prompts are, the more simple the drawing effect is. Also relatively general. The application is outstanding in creative design, anatomy art, original painting of game scenes, etc.! Appearance in game original paintings and scene applications: The style used by Deliberate can best be described as a fusion of digital art and realism. Character portraits are brightly colored and responsible, suitable for use as a reference for game character prototypes. The landscapes generated by Deliberate also look stunning, showcasing a range of shapes and textures, with realistic elements.

Here are a few accompanying versions of the model:


illustrate: The authors and users of the model are summarized

Negative prompts: (NSFW:1.4), (deformed, distorted, disfigured:1.3), poorly drawn, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs, (mutated hands and fingers:1.4), disconnected limbs, mutation, mutated, ugly, disgusting, blurry, amputation


Features : RPG game style, oil painting texture

As RPG is named, the model is based on the model of RPG character portraits trained by Dreambooth, including "Baldur's Gate", "Dungeon and Dragon", "Icewind Dale" series, etc., and more modern style characters. The currently released versions include V2 Beta, V3 Candidate 16, and V4. The author is currently optimizing V5, and V5 can be paid for early adopters.

The following uses and presentations are based on the V4 version, which has been trained to focus on attributes and elements of medieval and role-playing game characters, such as heavy armor, leather armor, mage cloaks and similar content. The model training was created using only original works including tools such as Unreal Engine, Custom Editing, IMG2IMG Paint and MidJourney. Stable Diffusion understands artist names or styles, so including them will still have an effect when using this model.

  • Clothing: RPG V4 has been trained on many clothing and fabric details. The current version leans heavily on floral-like motifs (flowers, leaves, and feathers), using RPG clothing types with cloaks and delicate fabric leather and stitched heavy armor and metal.
  • Range of styles: supports a large number of shadow styles and effects



Combining the author's recommended parameters and actual scene conclusions

  • Sampler: DPM++ 2S a Karras
  • Steps: 20~40, 50/75/150
  • Size:512*768
  • CFG scale: 3/4.5/6
  • VAE: vae-ft-mse-840000-ema
  • Negative prompts:

(flower:1.2), (Facial Marking:1.1), nude, (bad art, low detail, pencil drawing:1.4), (plain background, grainy, lowquality, mutated hands and fingers:1.4), (watermark, thin lines: 1.2), (deformed, signature:1.2), (blurry, ugly, bad anatomy, extra limbs, undersaturated, low resolution), disfigured, deformations, out of frame, amputee, bad proportions, extra limb, missing limbs, distortion, floating limbs, out of frame, poorly drawn face, poorly drawn hands, text, malformed, missing fingers, cropped

Cheese Daddy's Landscapes mix

Features : dream scene, plant environment

The Cheese model is also a mixed large model. There are currently seven versions: V1, V2 Delibrate added, V2 pruned, V3, V3.5, V4, and V4.1. The differences and parameters of different versions are introduced below.

The V1 hybrid model has:

  1. Roboetics
  2. Inkpunk
  3. DreamShaper
  4. Chroma V5
  5. H&A's 3DKX 1.1
  6. Analog diffusion
  7. The Ally's Mix
  8. Tokens: Analog style, chromav5, nvinkpunkChromav5 Keywords: Chromatic Aberration; Geometric Shapes; Bokeh; Depth of Field; Photorealistic; Cosmic;


Added fusion Deliberat

V3: Added AnythingV3 VAE: kl-f8-anime2LoRA: Jordan 3 , dpep2 768 (weight 0.8)

V3.5: Increased fusion with Pastel mix

V4.0: add the fusion of deliberate 2.0 , add the fusion of rev animated manipulative plug-in basil_mix

V4.1 (experimental) Deleted rev_animated and added a fusion of realistic vision

Clip skip: 1

  1. Size: (512x768 / 768x512) or (512x512)
  2. Sampler: DPM++ 2M
  3. CFG: 5-7
  4. Prompt: award winning photography, Bokeh, Depth of Field, HDR, bloom, Chromatic Aberration ,Photorealistic, extremely detailed, trending on artstation, trending on CGsociety, Intricate, High Detail, dramatic, art by midjourney, masterpiece, best quality, high quality ,extremely detailed CG unity 8k wallpaper


Fantasy World

Features : fantasy scene, partial Japanese manga style, pastel effect, with a little reality

The author uses Dpepteahand3 as the basic model for integration, the key tagger is Concept Artists, and the rest of the prompt words are determined according to his own drawing needs, mainly to remove non-scenery words such as human beings.

use :

Description :

The author recommends using parameters

  • Sampler: DPM++ 2S a Karras
  • Steps: 20
  • CFG scale: 7
  • Negative prompts:

lowres, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artist name, humans, castle, 1girl, solo, female

A-Zovya RPG Artist Tools

Features : RPG game adaptation

The author stated that this is a model that takes professional RPG industry artists into consideration. More creative inspiration for those working on video games, board and tabletop games, as well as concept art and book covers. The following is an introduction to the characteristics of different versions of the model:

Based on SD1.5 training——

  • V3 stronger painting style. High contrast and sharpness, more RPG knowledge.
  • V3 inpaint enhanced version
  • V2 High contrast and sharpness
  • The addition of V2 offset noise makes the contrast stronger and the image more realistic
  • V2 art training model, artistic, less details and more strokes, AI painting digital style is better
  • V2 repainted version of inpaint V2
  • V1 smoother rendering, simple painting effect
  • Repainted version of V1 inpaint V1

Based on SD2.1 training - SD 2.1 768 V1 Strong painting style, very coordinated with hands and objects, higher native resolution and details; it is not easy to produce nude pictures, and the prompt words are not very effective, but if you like to use LoRA Instead, here's LoRA .



The author recommends using parameters

  • Sampler: DPM++ 2S a Karras
  • Steps: 20-25
  • CFG scale: 5-8

Dungeons N Waifu's

Features: semi-realistic to realistic illustrations, can support the recognition of artist prompt words

"Dragon and Dungeon" game fusion model currently has four versions: 1.0, 1.6, 2.0FULL, and 2.6FULL. Versions 1.0 and 1.6 are more suitable for SD newcomers. Versions 2.0 and 2.2 have higher requirements for prompt word descriptions, but add shadows Effect. The author integrates the game fantasy world, character and creature design into this model, and users can create any fantasy art, character concept, portrait, design, and scene. In addition to painting, you can also do design work, vector illustration, logos, and a little animation.



The author recommends using parameters

  • Step: 20-70
  • CFG: 7-13.5
  • Sampler: latent (nearest or exact) upscale / denoise 0.4-0.7
  • Upscale by: 1.5-2
  • txt2img picture to inpaint denoise strength: 0.5-0.57
  • Whole picture inpaint: 0.53-0.56
  • only masked inpaint: 0.54-0.57
  • VAE: ft-mse-840000-ema-pruned.ckpt , kl-f8-anime2.ckpt

Download and try such a cool model, maybe you can also make an RPG game~



no dataCoffee time! Feel free to comment