Transcription
Listen to my advice!! Immediately stop subscribing to the AI tools you have in hand. Stop paying for those one-off AI chats. Because Google has quietly built a powerful and completely free AI productivity tool ecosystem that can help ordinary people easily build their own AI employee teams. However, 99% of people may not yet understand how to unleash its true potential. In today's content, I will thoroughly reveal the key functions within Google AI, completely igniting the synergistic potential of Google Gemini / Notebook LM, Opal, and Nano Banana / Veo3 YouTube Studio. After watching today's video, even if you are a complete beginner, you can easily have a YouTube script expert who understands your style, mass-produce hit script content, an excellent cover visual designer, and a 7x24 hour email marketing copywriter expert to help you convert potential customers into actual sales. You can even create an entire AI automated workflow with just one sentence, and it won't cost you a penny. Are you tempted? Let's get started!! Before we begin, you must first understand why it's Google AI. The answer is simple: Google's product ecosystem and powerful multimodal capabilities. You should know that there are very few products worldwide with over a billion users. However, Google alone has no less than ten. Imagine Google Search / Gmail / Google Docs, Android / YouTube. These platforms are now all connected by a super brain, Gemini. And the core advantage of all this is that it's free. Google is currently using this unbeatable free ecosystem to launch a dimensional attack on all expensive AI tools. So, are you ready to use Google AI to build your one-person company team? In the past, to build a successful one-person company, you might have had to force yourself to become a superhero: a content director who understands algorithms, a senior copywriter who can write hit scripts, a marketing assistant who replies to emails 24/7, and so on. But today, I can tell you that all these roles, the AI efficiency we often hear others talk about in the past few years, is actually the biggest lie. This is because you think you are improving efficiency by using AI in your daily life, but in reality, you are training new employees every day. This is not efficiency, but high repetitive communication costs. And Gemini's latest powerful function, GEM, is designed to solve this fundamental pain point. It's not a chat room, but your AI employee recruitment department. So, without further ado, let me quickly familiarize everyone with Google AI's core product, Gemini, and its usage tips. Let's get practical and recruit our four core AI employees. First, we can go to Google's backend. Here, there is a very obvious Gemini chat box. We can upload files for data input, or we can choose the scenario tools we use. For example, we can use Veo for video generation, Imagen for image generation, and Canvas for more complex content output with higher format requirements. The learning and tutoring function can help us quickly learn any knowledge point. The deep research function can help us conduct more in-depth and complex research work. In the bottom right corner, we can choose different Gemini models. For example, 2.5 Flash provides comprehensive assistance more quickly. And Gemini's 2.5 Pro model can provide deep reasoning capabilities and handle more complex mathematical or coding tasks. Here, for example, we can ask AI to generate a bedtime story picture book about a little fox. Then we can input the corresponding prompt and select Canvas in the tools, then click enter. You can see that after using the Canvas function, our page will automatically transform. The Canvas function here actually has multiple output methods and formats. For example, we can choose web output, infographic output, quizzes, flashcards, and audio summaries. If we continue to input the prompt for a picture book, you can see that a complete and very high-quality bedtime story picture book about a little fox has already been generated on the right side. The quality of this story picture book is very high, both in its textual expression and the drawing of the pictures on the right. After we generate the picture book, there is a very powerful function that allows us to directly export it to Google Slides. We can further edit it in Google Slides. If you don't need to export it to Google Slides, you can also directly click download or share your Canvas with other team members. Of course, besides the Canvas function, we can also use Gemini's deep research function for more in-depth business analysis or work-related research. For example, we can give a prompt to Gemini to break down the website of an independent station, its traffic sources, conversion keys, etc. At this time, Gemini's deep research will have specific directions for research and analysis. If this analysis direction does not meet our needs, we can modify the plan here. If we think it's okay, we can click start research or input start research. At this time, we just need to wait for a while and have a cup of coffee. After it's generated, we will get a deep breakdown report about bleame.com, including the structure of the entire sales funnel and the specific strategies it uses. This in-depth breakdown report also provides a very convenient table of contents. We just need to look at this table of contents, click on the parts we are more interested in, and then understand them. Besides the Deep research and Canvas functions, a very common and frequently used function in Gemini is using the image generation or video generation function. For example, let's test Gemini's video generation function. We can give a very simple prompt and see how the generated effect is. You can see that it generates a scene of a little fox lying in bed, preparing to fall asleep, and the entire scene is in a 3D animation style, looking very exquisite and dreamy. This is Google's very powerful video generation capability. Of course, currently, Google's Veo3 generation has a certain quantity limit and does not support free generation for everyone. If you want to use Google AI's image and video generation functions, it is more recommended to visit Flow or Whisk websites. For example, go to Flow's official website and click create a new project. In this project, we can create our video. Because an 8-second short film cannot be a complete video, we need to manage it as a project to create our short story film. For example, here we can generate video from text, use images to generate video, and use multiple materials to generate video. The menu in Flow allows us to set the size of the generated video. In Gemini, you may only be able to generate a horizontal long video, and the number of outputs for each prompt, as well as the model we choose. If you want to use Google's image generation function and want to generate images with more consistent characters or styles, you can go to the Whisk backend tool. It also supports setting various image sizes, inputting multiple image references, etc. This tool is also developed by Google. Here, you can also click to add animation effects to the images you generate, turning them into a short video of a few seconds. We only need to input simple video prompts, and it will generate a new video based on our video prompts and this image reference. It is important to note that we have already used up our video generation quota in Gemini, but here in Whisk, we still have 50 generation quotas. In Flow, we can also have more video generation quotas. How convenient is that? These are just the more conventional operations of Google AI. Next, let's talk about more advanced Google AI usage tips. To create a hit piece of content, the key lies in topic selection and strategy, not in directly starting to write the script. The first employee we need is our AI content director. They need to be able to break down the secret formula of competitor videos and help us research more excellent topic ideas. This requires using Gemini's most powerful and underestimated capability: directly analyzing YouTube video links. Let's go directly to Gemini's backend, click "Explore Gemini" on the left. Here, we can manage and create our AI employees. These are the Gems I have already generated. If we need to create a new one, we can click "Create New Gem" here. In the name field, we name our AI employee and describe its purpose. In the instructions, we can input our corresponding descriptive prompts. This prompt is essentially our AI employee's job description. At the same time, we will give our AI employee a very specific work goal, such as deeply analyzing the hit DNA of one or more successful scripts. Its specific workflow is within our AI employee's job description. We need to include a specific work SOP so that our AI employee can strictly follow our required steps to complete the task. In addition to specific work steps, we also need to provide more specific output constraints and requirements to make the content generated by AI more aligned with our needs. For example, while watching videos, we see a video like this. This video was posted a day ago and already has 160,000 views. Its topic is related to drugs in the US. We can copy the video link and go directly to Gemini's backend. We input this YouTube video and give it to our AI content director for breakdown and analysis. Of course, sometimes you will find that even if you provide a YouTube video link, it cannot be broken down well. This is because Gemini sometimes experiences service instability. In this case, you need to upload the entire video or the entire video script to Gemini for analysis. Then it is not limited to the YouTube platform; it can break down any video content. Let's carefully examine its analysis results. First, in the opening part, there is a crucial keyword: extremely high information density. The opening is a dual impact of vision and hearing. We can look at its opening. You can see that its opening description is very accurate: fast-paced editing. This opening is a very, very important factor in why the video went viral. Coupled with the structure of exploration and interviews, the entire rhythm is very well-paced. The ending part transitions to value sublimation, then summarizes and sublimates the theme, and sets up an interactive CTA task. Here, it will also specifically break down, for example, what hooks and techniques are specifically used in the video. We won't specifically discuss the content of this video; the key is to understand why this video went viral and the reason for its popularity. In addition to hooks, it also analyzes the video's value return points, i.e., which key points are very attractive to viewers. At the same time, there is also audience psychological analysis, such as the core emotional curve. Throughout the entire ten-plus-minute video, the audience's emotions are constantly stirred up in this way, generating many emotions, making people continue to watch. A particularly interesting point is that we will also see an analysis of video algorithm friendliness here. Why is this important? Because excellent creators know that our videos are not just produced for viewers but also need to comply with the platform's algorithm rules and cater to the platform's algorithms. Therefore, this analysis of algorithm friendliness is crucial. For example, the insertion of keywords, emotional resonance to promote interaction and comments, and duration design. This overall analysis is very complete. You can see that this entire YouTube analysis, I think, surpasses 90% of creators, even 95% of creators cannot achieve such detailed and pain-point-hitting analysis results. In the past, we relied on our own or our team's experience and accumulation of internet intuition to create content. But now, AI can completely become your very scientific and efficient hit content director, truly helping you discover hit topics. With such analysis conclusions, we can then ask the content director to further extract the hit DNA and successful hit formula. We can click continue or input continue. You can see that our content director has already given us a great summary of the hit formula. For example, based on the in-depth analysis in the first step, what kind of hit formula can be summarized? Right? The specific breakdown and detailed explanation of this hit formula will also be clearly explained to us. After we complete the analysis, we can click continue to let our AI content director generate a more specific strategy blueprint and optimization suggestions. With these optimization suggestions and understanding our hit formula, we can proceed to the next step, which is to create the specific script. At this time, we need to start recruiting our second AI employee, our AI script expert. We can go to the left sidebar, click on our "AI Top Scriptwriter Expert." After clicking edit, you can see its AI employee name, description, and specific instructions. This specific instruction is the same as the instruction structure for our AI content director. For example, role setting, core goals, work steps and SOP, output content format constraints, key point notes, and specific output requirements. At the same time, in the knowledge section, we can upload our previously generated scripts or hit script content. With these script references, AI can better understand our content style and produce scripts that better meet our needs and align with our past writing and content creation styles. The specific usage is also very simple. We can copy the analysis conclusion from our AI content director and input such a prompt. At this time, you can see that it will provide some new suggestions based on the hit knowledge base we provided and the hit logic of the benchmark video we just provided. Because the scripts in the knowledge base we provided to the AI employee are mainly AI-related topics, you can see that the new 10 topic directions are all more related to AI. For example, the first topic: "Victims and Reborns of AI Agents." This perspective combines social reality and the impact of AI to talk about the social reality and impact brought by AI agents to people. It's quite interesting. The second topic provided to us is "AI Hallucinations: Exploring the Dark Side." Delve into the dark side of AI. For example, if we intentionally feed AI incorrect information, do its feedback horror stories reveal the true dangers of intelligence? Telling about unknown dark social realities is also a type of topic that easily goes viral. Let's assume we choose the first topic. It will quickly start to outline the corresponding script content based on our chosen topic. For example, it starts with an analysis of the topic's attractiveness, algorithm positioning, digging into core controversies, and then constructing an outline based on the PAS structure. For example, the topic it gives us is: "In the Era of AI Agents, are we destined to be eliminated outsiders or reborn individuals who master new power?" What is the dividing line between these two completely different destinies? You can see that this topic is also quite interesting. For interview-style content or content exploring social reality, it's a very good entry point. If we think it's okay, we can click continue. Then we can see the opening it generates: "What you are seeing now is not a movie scene. AI has its first hands and feet." What do you think of this script opening? I think this script opening is excellent. Then we can continue with the specific script content creation. Because in the AI employee's SOP prompt, we specifically asked it to generate step by step to ensure that the script for this long video is of very high quality, has sufficient information density, and meets our needs. For a hit YouTube video, you have a hit topic, the corresponding script and outline, and even specific complete script content. What else is crucial? Your cover. We need to rely on the cover to get viewers to click on our videos. The third employee we need to recruit is our AI visual designer. They will use Google's Nano Banana and other powerful image generation models to generate corresponding video thumbnails and covers for us. We can input such a prompt, copy and paste the topic we just determined, and ask it to provide specific references. Of course, we can also provide corresponding references by adding files and uploading images. This might even be better. Let's first look at some cover ideas provided by our cover visual designer. You can see that the topic ideas are very bold and interesting. You can already imagine the scenes in your mind. Many people think AI has no creativity, but perhaps it's just that you haven't used AI well! Here you can see the prompt our AI visual designer gave us for image generation. We can copy and paste this prompt directly into the Whisk backend, input our prompt, and select the desired image size, which is a horizontal size, and then click enter. You can see that this video cover has been created. This video cover is quite impactful. We just need to add some specific text for the video. For example, we can copy this video and paste it into Canva for specific text editing. If you are generating an English video, you can directly ask AI to generate it. Because Whisk or Google AI has better support for text generation within images for English, the support for Chinese is relatively weaker. Okay, now we have the topic, the content, and the cover. The final step is to do some promotion. Let's assume we have many private traffic customers or many viewers who have joined our private traffic. We need to promote this video or conduct marketing activities in our private traffic, such as our WeChat official account, our emails, and other places. Then we need to recruit our fourth employee, who is responsible for building our email marketing funnel. We also go to "Explore Gemini" here. We can see there is an "AI Top Email Marketing Copywriter Expert" here. We click on it. Here, for example, we can randomly copy a video and ask AI to write an email copy for this video. Let's look at the 10 email subject lines our AI employee gave us. You can see that the titles and opening hooks are very attractive. The first one is: "$5 drugs and $8 lunch. On this street in Philadelphia, it's cheaper to destroy yourself than to eat a meal. We went to the scene." The second one is: "I'm in America's zombie city. This is not a movie scene. The people on the street are real, and the police are watching." There are a total of 10 angles. The first few angles are quite good. Let's use the first one, for example. After confirming the selection, it will start to create and analyze the email script outline. We don't need to look at it too closely here. If you are interested, you can take a look. If not, you can click continue. Then we can see the subject is "5 dollars for drugs and 8 dollars for lunch." Preview text: When destroying yourself is cheaper than staying alive, what will this place become? It becomes a street of zombies, but the most chilling thing is that the police are in police cars just a few meters away, watching it all. The law seems to be ineffective here. What do you think of this copywriting ability? To be honest, I have hired many people for copywriting positions in the past. This copywriting ability may have surpassed more than 80% of copywriters we interviewed. Of course, we can also recruit more AI employees. For example, we can recruit AI content specialists, i.e., AI employees for multi-platform content distribution. We can repurpose our YouTube long videos into hit short videos, WeChat official account copy, Line/IG post copy, Thread posts, etc. It's all perfectly fine. Now our AI team has a decision-maker, a copywriter, and a visual designer. But this is not enough. Especially when facing more complex tasks, we sometimes still need to rely on the middle office of the AI team, which is our Google AI Studio. You can search for the keyword "Google AI Studio" and click on it. First, here we can choose from four entry points. The first is "Try Nano Banana," we can choose Google's latest Nano Rubber model. We can choose Veo3.1, the latest video generation model. We can choose "text to speech," where we can generate voiceovers. Under the "home" menu, there is a chat box entry. This entry is closer to Gemini, but it differs from Gemini in that we can directly choose more rich Google large models in the top right corner, such as Google's Nano Banana, Google's Gemini 2.5 Pro, Google's Flash-lite, Google's Imagen 4, or Imagen 4 Ultra. What is the difference between Imagen 4 and Nano Banana? Imagen 4 has higher quality in text rendering. Of course, Imagen 4's text rendering is more for English text, and its performance is more prominent. The biggest technical bottleneck in AI image generation is garbled text. For example, when we ask it to add a title in AI image generation, it always gives us a bunch of alien text. But here in AI Studio, we can choose Google's latest Imagen 4 model. In addition to choosing models, we can also customize what requirements the AI output content needs to meet in "system instruction," or customize the style of AI output content. At the same time, in "template," we can choose the creativity of the generated AI and the image generation size. Simply put, Google AI Studio gives us more freedom and allows for more customized settings. Compared to Gemini, its operation complexity is slightly higher, but not much. In addition to the specific usage tips I have introduced in Gemini, a special feature of AI Studio is that we can directly use Google AI Studio to generate speech. In "home," we can choose "text to speech with Gemini." Here, we can generate the AI voice we want. For example, in "Raw structure," we can specify the general scenario of voice generation, participating roles, etc. In "script builder," we can specify "style instruction," i.e., the tone and intonation of reading, and add different speakers with their different styles. If we don't need multiple speakers, we can choose "single speaker" in "mode" on the right. If multi-speaker dialogue is needed, choose "Multi speaker Audio" here. At the same time, we can also choose a more specific speaker's timbre in "voice." Here, let's demonstrate specifically. We can first choose "single speaker Audio." Here we can simply set the style prompts for generating speech, such as a shocked, fast-paced speaking style. In "text," we input the specific script content we want AI to read. We can give the script opening we just generated in Gemini to the AI and see its specific generation effect. After the above operations, we now have an on-demand execution team. But the key problem is that your team is still forgetful. They only understand general skills, not your business, not your products, not your SOP details, and they don't have the shared memory of a company team. So, the next step is to establish our one-person company's shared brain, which is our one-person company knowledge base. And the employee responsible for managing this brain is our Chief Knowledge Officer, Notebook LM. Some people might think, "Gemini can already upload files, why do we still need to use Notebook LM?" There is a fundamental difference here. Gemini just looked at your files and had a brief memory, but Notebook LM truly becomes your team's knowledge base. When you ask Gemini a question, it will mix knowledge from the internet with the materials you uploaded to answer you. But when you ask Notebook LM a question, it will 100% answer based only on the data you fed it. It will not hallucinate, guess, or cite any external information. This is very important for a company's brain. You certainly wouldn't want your team's knowledge base and SOPs to contain a lot of hallucinated answers or inaccurate information. Let's go directly to the official website of Notebook LM. Here we can see that it has already provided us with some examples. We have different knowledge bases. Here, we can choose to build our own knowledge base. Click "New." Here, we can add data sources. It provides many ways to add them. For demonstration purposes, let's just give a YouTube link. You can see that it quickly absorbed the content of this video. For example, it extracts a video from the YouTube channel. The content focuses on the serious drug problem in Philadelphia, USA. If you run your own YouTube channel, you can give Notebook LM all the video links from your YouTube channel. We can also upload up to 50 documents and sources at once. Based on the knowledge sources we added, or by uploading your team's SOP documents, we can then have AI perform information visualization. For example, we can ask it to generate a corresponding mind map to understand the specific content explained in this video. In addition to generating mind maps, you can see that Notebook LM also provides many other options, such as audio summaries and video summaries. We can use the video summary function to generate a demonstration video explained by AI. You can see the specific generation effect. How quickly it helps us generate a brand new video content based on the original topic and content. In addition to the video summary function, there is also an audio summary function that we can test. First, let's look at the Vibe English channel. This channel has only posted 69 works but has already gained 580,000 subscribers. This subscription data is very impressive. Let's also look at its main content, which is English listening content that lasts for over an hour to help everyone practice English. In addition to this type of English listening content, it also released an English conversation podcast content four months ago. Let's listen to it. If we want to generate corresponding content, we can directly go to the Notebook LM backend, click "Add Source," select "YouTube Link," paste our YouTube link, and click "Insert." At this time, it will have corresponding output content. Then we click "Audio Summary," and it will generate a new AI podcast content based on the content and topic we provided. With Notebook LM, we now have an all-knowing shared brain and a versatile execution team. But up to this point, we ourselves are still the intermediary. We must manually copy and paste insights from Notebook LM to Gemini's AI employees, or manually copy and paste Gemini's scripts into AI Studio for voice generation. We still cannot truly achieve complete hands-off operation. So, we still need a tool that can connect all the execution parts and AI employees, an automated workflow tool, or a management role called an AI manager. This is Google's latest launch this year, Opal, which is still in beta testing and only available in some countries. If you are not in a country where Opal is available, it will prompt: "(Opal is not yet available in your country/region)." You can switch your IP address to the United States to log in and use it. Compared to the AI employees mentioned at the beginning, it is more like a disciplined, tireless AI manager. Its core task is to turn your tedious manual copy-pasting work into a fully automated content production line, which is the landing form of AI agents. Let's do a practical demonstration. After coming to the Opal backend, you can see that it has already provided many examples. If we want to build our own Opal workflow, we can click "Create New" here. Then we will come to such a backend. This backend is divided into several parts. One is "user input," which is our input item, our trigger. "Generate" is the intermediate node. For example, after inputting some content, each of these clicks will generate a corresponding node. Each node is an execution, a factory assembly line node. At the same time, we can also add our own knowledge materials and content. If you are familiar with N8N or Coze, you should not be too unfamiliar with such a working interface. Compared to workflow tools like Opal, N8, and Coze, the biggest difference is that we can generate a complete AI workflow with just one sentence prompt. Let's assume we generate such a workflow: it can convert a YouTube video into a blog post. You can see that after clicking enter, AI has automatically built this workflow for us. You can see that this workflow has a total of 3 nodes. The first node is that we need to input a YouTube URL. The second node here is to help us generate an SEO-optimized blog post. The third node is to help us adjust and modify the format of the blog post to better meet our needs. After generating the corresponding workflow, we can go to the interface on the right. In "Preview," we can have conversations and interactions with our workflow. In the "console" here, we can see the specific steps of the workflow: step 1, step 2, and step 3. In "step," we can see the specific prompts. In "Theme," we can see the theme style we want to switch to, for example, the theme style of the preview interface. Now we can try the workflow's running effect. We click "start" and input the YouTube video for which we want to generate a blog post. Now our blog post has been generated. You can see the specific webpage. It not only generates articles but directly gives us a blog post website. It has already generated the HTML file and code for the webpage. Of course, because the prompt we generated this time is relatively simple, and the workflow we actually generated is also relatively simple, with only three nodes, you can see that the webpage and articles generated are relatively basic. Next, we just need to continuously interact with AI in this interface, using natural language conversations to let it help us optimize this AI workflow, for example, optimizing specific copy, giving better copy prompts, letting it optimize the visual effect of our webpage, making it generate more aesthetically pleasing webpage styles, all of which are possible. Of course, we can also add assets to give our company or our team more internal materials, and the final generated copy will be more aligned with our needs. Besides creating blog posts/SEO articles by converting YouTube videos, we can also achieve various AI workflows, such as converting YouTube videos into short video scripts, converting YouTube videos into your Thread posts, converting YouTube videos into your IG posts, etc. We can even directly ask Opal to generate a complete AI video for us based on our prompts and selected content themes. You can see that in the "generate" node here, we can choose Gemini, different multimodal models, such as Imagen 4 image generation model, AudioLM audio generation model, Veo2/Veo3 video generation model, and Lyria2, which you may be less familiar with, a music generation model. So you can see that Opal supports Google's multimodal capabilities, including text, audio, video, and images. With such complete multimodal capabilities, we can truly build an AI video AI workflow. If time permits in the next issue, we can also demonstrate it practically. With Gemini/Notebook LM, Google's AI Studio, Flow / Whisk, and the newly launched Opal AI Manager, your AI employee team truly has autonomous action. You are no longer just a creator, but an owner of a system. The truth of AI efficiency lies not in how many AI tools you use, but in how many processes you link and whether you truly hand over your critical business workflows to AI. The core key to using AI is to stop constantly training new employees and repeatedly communicating with AI from today onwards. Start learning to manage your AI employee team and build your own business system.