Background
With the rise of language models for generating images from text, SolidUI aims to help people quickly build visualization tools. The visualization content includes 2D, 3D, and 3D scenes, thus quickly constructing three-dimensional data demonstration scenes. SolidUI is an innovative project that aims to combine natural language processing (NLP) with computer graphics to achieve text-to-graphic functions. By building its own text-to-graphic language model, SolidUI uses the RLHF (Reinforcement Learning Human Feedback) process to implement the process from text description to graphic generation.
Project link: https://github.com/CloudOrc/SolidUI
Project mirror link: https://gitee.com/CloudOrc/SolidUI
Community official website: https://website.solidui.top
Official website project address：https://github.com/CloudOrc/SolidUI-Website
Why is there a need for a graphic persona?
The essence of a persona is effective persuasion, guiding GPT to focus on solving problems in a particular domain. For example, if you are a data analyst and want to use GPT to generate graphics, the design concept is also based on the design and extension of the persona.
How to create a graphic persona?
Here is a 3-step strategy for creating a graphic persona:
Step 1: Establish the persona (WHO)
The underlying logic of establishing a persona is to have GPT play a specific role, focusing on providing solutions to related issues within the corresponding professional field. In this step, we need to determine GPT's role and professional field.
For example, we can have GPT play the role of a "professional data scientist and data analyst", focusing on providing solutions to problems related to graphic generation.
Step 2: List the requirements (HOW)
Once the persona is established, we need to list the related requirements, clarifying what GPT should do. For example, we can require GPT to generate corresponding graphic materials based on the given keywords.
Step 3: Give commands (WHAT)
The last step is to give specific commands so that GPT knows what we need. For example, we can give such a command: "Generate a simple graphic"
Through these 3 steps, we can construct a graphic persona, allowing GPT to generate corresponding graphic materials according to our needs.
How to reuse a graphic persona?
In actual use, we may need to use different graphic personas based on different needs. At this time, we can reuse existing personas through some simple operations.
For example, we can save an existing persona as a template, and then make some minor modifications when needed, to reuse this persona.
In addition, we can use some techniques to improve the efficiency of the persona and the quality of the output. For example, we can add a sentence "Please think step by step" to GPT's input, so that GPT will think more carefully about each step, thereby improving the quality of the output.
SolidUI Model Proxy
The model proxy can dynamically load different types of models, which are based on personas, receive user input, and generate visual class resources (e.g., python/html, etc.) after user input, rendering any graphics. The specific source code is in the soliduimodelui module.
How to Become a Contributor
Here are some ways to contribute to the SolidUI community.
Contribute to Official Documentation: Discover deficiencies in the documentation, optimize the documentation, and continually update the documentation to contribute to the community. By contributing to the documentation, developers can familiarize themselves with how to submit PRs and truly participate in community building. Reference strategy: https://github.com/CloudOrc/SolidUI/discussions/54
Contribute Code: We have combed through simple and easy-to-start tasks in the community, which are very suitable for newcomers to contribute code. Please check the newcomer task list: https://github.com/CloudOrc/SolidUI/issues/12
Content Contribution: Publish content related to SolidUI open source components, including but not limited to installation and deployment tutorials, usage experience, case practices, etc. The form is not limited, please contribute to the assistant. For example: https://github.com/CloudOrc/SolidUI/issues/10
Community Q&A: Actively answer questions in the community, share technology, help developers solve problems, etc.
Others: Actively participate in community activities, become a community volunteer, help community publicity, provide effective suggestions for community development, etc.
Top comments (0)