Acertainmodel

🚀 ACertainModel

ACertainModel is a latent diffusion model designed for anime enthusiasts. It can generate high - quality, detailed anime - style pictures with just a few prompts and supports danbooru tags.

🚀 Quick Start

Try full functions with Google Colab free T4

Check Twitter #ACertainModel for community artworks

✨ Features

• High - Quality Generation: This model is capable of producing high - quality, highly detailed anime - style pictures with only a few prompts.
• Danbooru Tags Support: Similar to other anime - style Stable Diffusion models, it supports danbooru tags, including artist names, for image generation.
• Advanced Training: Through a series of fine - tuning and training methods, it has better performance on details such as eyes and hands.

🔧 Technical Details

Since the laion - aesthetics introduced in the Stable - Diffusion - v - 1 - 4 checkpoint hindered finetuning anime style illustration generation model, Dreambooth was used to finetune some tags separately to make it closer to what it was in SD1.2. To avoid overfitting and possible language drift, a huge amount of auto - generated pictures from a single word prompt were added to the training set, using models popular in the community such as Anything - 3.0, along with partially manual selected full - danbooru images within a year, for further native training.

The method of LoRA, which finetunes the attention layer solely, was also considered to have better performance on eyes, hands, and other details.

For copyright compliance and technical experiment, it was trained from few artist images directly. It was trained on Dreambooth with pictures generated from several popular diffusion models in the community. The checkpoint was initialized with the weights of a Stable Diffusion Model and subsequently fine - tuned for 2K GPU hours on V100 32GB and 600 GPU hours on A100 40GB at 512P dynamic aspect ratio resolution with a certain ratio of unsupervised auto - generated images from several popular diffusion models in the community with some Textual Inversions and Hypernetworks. No tricks on xformers and 8 - bit optimization were used for better quality and stability. Up to 15 branches were trained simultaneously, cherry - picking about every 20,000 steps.

💻 Usage Examples

Basic Usage

from diffusers import StableDiffusionPipeline
import torch

model_id = "JosephusCheung/ACertainModel"
branch_name= "main"

pipe = StableDiffusionPipeline.from_pretrained(model_id, revision=branch_name, torch_dtype=torch.float16)
pipe = pipe.to("cuda")

prompt = "pikachu"
image = pipe(prompt).images[0]

image.save("./pikachu.png")

Advanced Usage

When using the Hosted inference API for online preview or generating images with this model, parameters are not allowed to be modified. It seems that it is generated with Clip skip: 1. For better performance, it is strongly recommended to use Clip skip: 2 instead.

Here is an example of inference settings, if it is applicable on your own server: Steps: 28, Sampler: Euler a, CFG scale: 11, Clip skip: 2.

📚 Documentation

This model can be used just like any other Stable Diffusion model. For more information, please have a look at the Stable Diffusion.

You can also export the model to ONNX, MPS and/or FLAX/JAX.

📄 License

This model is open access and available to all, with a CreativeML OpenRAIL - M license further specifying rights and usage. The CreativeML OpenRAIL License specifies:

1. You can't use the model to deliberately produce nor share illegal or harmful outputs or content
2. The authors claims no rights on the outputs you generate, you are free to use them and are accountable for their use which must not go against the provisions set in the license
3. You may re - distribute the weights and use the model commercially and/or as a service. If you do, please be aware you have to include the same use restrictions as the ones in the license and share a copy of the CreativeML OpenRAIL - M to all your users (please read the license entirely and carefully) Please read the full license here

Examples

Below are some examples of images generated using this model, with better performance on framing and hand gestures, as well as moving objects, comparing to other analogues:

Anime Girl:

1girl, brown hair, green eyes, colorful, autumn, cumulonimbus clouds, lighting, blue sky, falling leaves, garden
Steps: 28, Sampler: Euler a, CFG scale: 11, Seed: 114514, Clip skip: 2

Anime Boy:

1boy, brown hair, green eyes, colorful, autumn, cumulonimbus clouds, lighting, blue sky, falling leaves, garden
Steps: 28, Sampler: Euler a, CFG scale: 11, Seed: 114514, Clip skip: 2

Is it a NovelAI based model? What is the relationship with SD1.2 and SD1.4?

See ASimilarityCalculatior