- Neta Noob V1.0 Release Notes
- I. Overview
- 1.1 Model name and version
- 1.2 Model Introduction
- II. Model features
- 2.1 Diverse styles
- 2.2 Great characters coverage
- 2.2 Stable high-quality output
- 2.3 Strong prompt-following
- III. User Guide (partially quoted from the NoobAI-XL User Manual)
- 3.1 Prerequisite
- 3.2 Download
- 3.3 Tips for Prompting
- Recommended quality words and negative words
- Examples
- IV. Parameter Settings
- 4.1 Sampler
- 4.2 Iteration step settings
- 4.3 CFG Scale Adjustment
- 4.4 The resolution
- V. Model testing
- 5.1 Test cases
- 5.2 Human Evaluation
- VI. Conclusion and Future Work
Neta Noob V1.0 Release Notes
Neta Noob V1.0 Release Notes
1月25日修改
I. Overview
1.1 Model name and version
NETA NOOB V1, optimized based on NoobAI-XL 178.
1.2 Model Introduction
This model relies on cutting-edge Deep learning technology and is specifically developed for the image creation needs of the ACGN domains. Whether it is for anime enthusiasts to create fan works, professional artists seeking inspiration to assist in drawing, or related merchandise industries for character design, scene drawing, etc., it can play a huge role. It is committed to accurately transforming the fantasy world in the user's mind into visually exquisite images. During the Model Training process, we used rich and diverse data resources, including dataset from Danbooru (8,400,000) and carefully hand-picked 400,000 pairs. With powerful computing power, we conducted 6 rounds of deep training with 8 x A100 to ensure that the model could learn various features and style essences of ACGN art styles.
