Update README.md

README.md

@@ -8,6 +8,27 @@ pinned: false
 license: mit
 ---
 
+## Update 2023.9.2
+
+Two recent developments:
+
+1. The **automated training process for `v1.4` has been deployed**, and the model's quality has improved significantly compared to before (for more technical details, see: https://civitai.com/articles/2064/2023-8-31-release-of-v14-training-automation-process). We are now in the process of thoroughly cleaning the dataset and retraining the model.
+
+2. We now support LoRA training for characters in anime videos, and the entire process is highly automated.
+
+Based on the second point, we have decided to **solicit anime requests from our users and train LoRA models for the characters** (currently limited to female characters with clearly defined human features). If you want a LoRA model for a character from an anime, please follow these steps:
+
+- **Join the CyberHarem organization** (the organization has no approval process and can be joined directly; LoRA requests from users who have not joined the organization will be given lower priority).
+- Create a discussion [here](https://huggingface.co/spaces/CyberHarem/README/discussions?status=open&type=discussion) and provide the following necessary information:
+  - **The official English full name of the anime** (if there is a commonly used English abbreviation, you can also provide it).
+  - A **magnet link for anime video resources that can be downloaded** normally (or other convenient ways), the video resources should have a **higher resolution** for better training quality, and it is advisable to use videos **without subtitles** (subtitles may have a negative impact on training quality).
+  - **The anime's wiki information page**, and it must include the English names and images of all main characters (this information will be used to distinguish characters).
+- The requester does not need to specify the names of the characters to be trained because **we will train all characters detected within that anime series who have a certain minimum threshold of images**. However, due to this approach, we may not be able to train characters with very low screen time that do not have a sufficient number of images available. We appreciate your understanding in this regard.
+
+**All three pieces of information are essential**; otherwise, the training process cannot proceed. After providing the above information in the discussion, if you see a reply saying "Request Accepted.", it means that the request has been accepted, and training will begin soon. When the model is trained and uploaded to CyberHarem, you will receive another reply in the discussion, which will include the model repository of the trained characters. **Requesters should provide a detailed evaluation of the model's quality and select a best step for each model as much as possible**. The models will then be published on civitai later with the requester-recommended best steps (if the requester does not provide further information, the best step determined by automatic evaluation will be used for publication).
+
+In addition, CyberHarem is a non-profit technical team that works purely out of interest, so **we do not charge any fees in any form**. However, our computing resources and team members' working time are limited, so **we cannot guarantee the delivery time of models in principle**. We will do our best to complete them as soon as possible under the circumstances, and we hope for your understanding in this regard.
+
 ## What is this?
 
 As you can see, this place is called `CyberHarem`, a centralized repository for anime waifu images dataset and LoRA models.