2024年AI到底从何学起？AI绘图、AI大语言模型、AI生成视频、AI生成语音各个赛道下都有哪些值得学习的项目 midjourney、SDwebUI、comfyUI、DALL-E3各有什么优势？

This paragraph discusses the recent surge in interest in AI following the release of Sora and the subsequent delisting of Dr. Li Yizhou's course. It highlights the competitive nature of the AI field and the general public's growing curiosity about AI technologies. The speaker refrains from commenting on the controversy surrounding Dr. Li but emphasizes the importance of providing structured educational content for newcomers to AI. The video's creation date is noted for context, and the rapid evolution of AI tools is acknowledged.

The speaker provides an in-depth overview of AI image generation tools, focusing on three主流 platforms: MJ, SD (Stable Diffusion), and DALL-E. Each tool's strengths and weaknesses are discussed, including user-friendliness, image quality, and controllability. The paragraph also touches on the commercial viability of using these tools and the challenges of achieving consistency in outputs. Various usage scenarios and the learning curve associated with each tool are also explored.

This section delves into the specifics of SD applications, namely WebUI, Fooocus, and ComfyUI. The speaker shares personal experiences and insights into the functionalities and user-friendliness of each application. WebUI's plugin ecosystem and model flexibility are praised, while Fooocus is recommended for beginners due to its simplicity and free usage. ComfyUI's node-based interface is acknowledged for its professional workflow customization capabilities, though it's noted to be more challenging for beginners.

The speaker introduces Adobe Photoshop's AI feature, Firefly, and its integration within the popular image editing software. The benefits of Firefly for users with Photoshop experience are highlighted, along with its role as an辅助 tool in the creative process. The paragraph also mentions the limitations of Firefly, such as the requirement for a正版Photoshop and potential instability in破解版软件. The speaker's personal recommendation for learning paths involving these tools is provided, emphasizing the importance of understanding each tool's parameters for effective use.

This paragraph covers the concept of large language models and their evolution into multimodal capabilities. The speaker explains the functionalities of ChatGPT and its multimodal upgrade, GPT4, and compares them to other AI language models like Gemini. The paragraph discusses the practical applications of these models, such as summarizing information, writing, and translation, and touches on the challenges of accessing paid versions for users in certain regions. The speaker also mentions emerging models like claude 3 and their potential impact on the AI landscape.

The speaker explores the field of AI-generated videos, noting the technical challenges and the hype surrounding platforms like Sora. The limitations of current video generation tools are acknowledged, and a warning is issued against fraudulent claims of Sora's capabilities. The paragraph also discusses AI voice synthesis, including text-to-speech and voice translation, highlighting the advancements in natural-sounding voice generation and the potential for realistic lip-syncing in videos. The speaker provides an overview of several AI voice projects, emphasizing the leading position of 11labs and the innovative efforts of companies like heygen.

In conclusion, the speaker reflects on the rapid development and diversity of AI technologies across various fields, including image generation, language models, video generation, and voice synthesis. The paragraph emphasizes the challenge of keeping up with the fast-paced evolution of AI and the multitude of projects available. The speaker expresses a desire to help viewers navigate the AI landscape and encourages engagement through comments and social media for the latest updates and discussions on AI advancements.