Checkpoint: Commonly used Stable Diffusion models recommendation

Checkpoint: Commonly used Stable Diffusion models recommendation

Civitai is a platform that allows people to easily share and discover AI artistic creation resources. Platform users can upload and share their own AI-generated custom models trained using their own data, or browse and download models created by other users. These models can then be used with AI art software to generate unique works of art.

Today we are going to introduce 12 commonly used checkpoint models for Stable Diffusion on Civitai.

Table of Contents

  1. Civitai model classification
  2. Checkpoint model recommendation
    1. DreamShaper
    2. Lyriel
    3. Protogen
    4. Dreamlike
    5. Realistic vision
    6. Counterfeit
    7. MeinaMix
    8. DDicon
    9. Product Design
    10. Isometric Future
    11. ReV Animated
    12. Architectural design sketches with markers

Civitai model classification

The models on CivitAI are mainly divided into four categories: Checkpoint, LoRA, Textual Inversion, and Hypernetwork, which correspond to 4 different training methods.


A large model obtained through Dreambooth training method. It is characterized by good rendering effect. However, because it is training a complete new model, the training speed is generally slow.

The generated model file is large, usually several G, and the file format is safetensors or ckpt

Checkpoint model recommendation

Model storage path: /stabl-diffusio-webui/models/Stabl-diffusion



  • Download address:
  • Model description: It is capable of a variety of styles and can achieve top-notch light and shadow effects and details of characters and scenery.
  • Recommended parameters: DPM++2M Karras, Clip skip 2 Sampler, Step25-35+


  • Download address:
  • Model description: Protogen is a fusion of many different excellent models and is suitable for many styles. Protogen v2.2 prefers digital painting (trigger words: modelshoot style); Protogen x3.4 prefers real people (trigger words: modelshoot style, analog style, mdrny-v4 style, nousr robot).
    In the Protogen model, the camera position and character movements can be precisely controlled through prompt words to generate character pictures from different angles and movements, for example: side (from_side:1.3), above (from_above:1.3), behind (from_behind:1.3), below (from_below:1.3), hand on head (hand_on_head:1.2), sitting (sitting:1.2)


Realistic vision




  • Download address:
  • Model description: A model that generates B-side style elements needs to be used in conjunction with the corresponding version of DDicon lora. The V1 version has more details, while the v2 version is more concise and can be used with the LoRA model DDICON_lora.

Product Design

Isometric Future

ReV Animated

Architectural design sketches with markers