The speaker opens with a direct promise: two tools you can use today in your marketing or your business. Both of them produce results that until very recently only a professional team could deliver. That framing matters because it sets the bar for what to expect, not incremental improvement but a category shift.

The first tool is ChatGPT ImageGen2. The second is Erwin, a platform that gives every student access to all the leading AI models in one place. The lesson covers both in sequence, starting with ImageGen2 and walking through a real workflow before moving to Erwin.

For a long time, ChatGPT used ImageGen 1.5. It was very good, but not quite good enough to be truly usable for finished marketing work. ImageGen2 changes that. The difference is described simply: it is pixel perfect and gives you exactly what you asked for. That precision is what makes it cross the line from a novelty into a practical production tool.

The phrase the speaker uses is important: pixel perfect, gives you exactly what you asked for. Previous versions would approximate. They would get the mood or the color range right but miss the text, the logo placement, or the layout detail. ImageGen2 closes that gap, which is why the workflow in this lesson is practical rather than theoretical.

The core use case is this: say you are promoting an event, a fundraiser, a banquet, a lunch, and there is marketing work out in the world that you admire, that you would love your team to produce. You can now hand the AI a reference and say, start from this, then make it your own. That instruction, start from this then make it your own, is the mental model for everything that follows.

The speaker demonstrates by taking a screenshot of an Instagram grid they admire, uploading it to ChatGPT, and asking it to replicate the look but rebrand it for CloudWise Academy. The result is that very same brand style, but it is a completely new image: pixel perfect, word perfect, and fully rebranded.

One clarification the speaker makes directly: you would not literally copy someone else's design and slap your name on it. The demonstration is for dramatic effect, to show what is possible. The actual workflow uses a reference for style inspiration, not duplication. The distinction matters both legally and creatively.

The first step is mechanical but important to get right. Open a free ChatGPT account and start a new chat. Then find a webpage, design, or image online that you admire. The speaker uses the Google Learning page as the reference in this walkthrough. Any page with a layout you want to learn from will work.

Once you have a reference, take a screenshot of the page. If you do not know how to take a screenshot, or if the image is embedded in a page rather than a standalone file, there is a simpler path: right-click any image online, save it to your computer, and then upload that saved image directly into ChatGPT. You do not need a screenshot tool. A saved image file works just as well.

The key principle here is that you are giving the model a visual reference, not just a text description. The difference between saying make me a hero section with a blue background and uploading an image you admire and saying match this is the difference between vague and precise. The image does the describing for you.

What makes Gen2 especially powerful is that it does not just generate images. It can research while it generates. That means your prompt can include a URL and the model will pull real content from that site before building the image. The speaker calls this Gen2 Research Mode.

The exact prompt structure the speaker recommends looks like this: Research cloudwiseacademy.com, then make a mock-up of a homepage hero section. Rewrite the content, rebrand the look and feel to match. Make the design as close to the reference as you possibly can. You adapt that structure to your own site and your own reference image.

To execute: click the plus button in the chat input, add your screenshot or saved image, then type your prompt. The speaker is clear that you should use your own words. The idea is to say: use this reference image, and then ask it to be as close or as different as you want it to be. Then click send and let it generate. The research step means the model is reading your actual website content, not guessing at what your brand says.

The first result may not be a complete mockup or exactly what you were looking for. That is expected and it is not a failure of the tool. The speaker is direct: you need to talk to it like it's your employee or your friend or your partner. Each follow-up prompt is a correction, not a restart. You are iterating, not starting over.

In the demonstration, the first result did not match the brand and included a subject the speaker does not teach: cloud skills. The follow-up prompt was specific: That's great, but use my logo, match my brand colors. We don't teach cloud skills. We teach AI for beginners, educators, and professionals. That one correction moved the output significantly closer. The colors matched. The content was accurate. The logo was still off.

The speaker makes an important point here about mindset: you're iterating very fast, so I don't want you to be disappointed if you're not getting what you want or you can't ultimately get the very final thing you want because you're getting so close. The goal of iteration is not perfection in one shot. It is narrowing the gap with each exchange. Each prompt you write is faster and cheaper than a revision cycle with a human designer.

When iteration through follow-up prompts gets you close but one element is still wrong, ImageGen2 has a built-in edit mode that lets you target a specific part of the image. In the demonstration, the logo area was not rendering correctly after multiple rounds of prompting. Rather than prompting again and hoping the whole image updates, the speaker clicks directly on the generated image to enter edit mode.

Inside edit mode, there is a paint tool. You use it to brush over the exact area you want to change, just as you would use a highlighter on paper. Then, in the text field below the image, you describe the fix. The speaker types: You didn't use my mark or my logo. Please fix that. Then uploads the logo image again so the model has a direct reference for what should appear in the painted area.

Once you click send, the model goes to work on only that painted region. The rest of the image stays intact. The result is a mockup you can hand your designer, drop in a flyer, paste into a slide, or use as a finished mockup. The speaker also notes: if you want different photos in it, give the photos and tell it to swap them out. The same edit-mode logic applies.

The second tool is Erwin. CloudWise Academy gives every student access to all the leading AI models in one place through Erwin. It looks and behaves like any AI you have used, but instead of being locked into one company, you choose your model. The interface is familiar. The difference is the model selector.

The three leading AI companies right now are OpenAI, Google, and Anthropic. Their flagship models, their top models, are GPT 5.4, Google Gemini 3.1 Pro, and the newest model from Claude, which is Opus 4.7. All three are available inside Erwin. You pick the model you want to work with and start chatting. You are working with the most capable AI models on the planet.

The practical significance of a single interface is that you do not need to manage three separate accounts, three separate billing plans, or three separate chat histories. You switch models the way you would switch tabs. Different models have different strengths. Having access to all three means you can match the model to the task rather than being stuck with whatever model you pay for that month.

To use GPT 5.4, Google Gemini 3.1 Pro, and Claude Opus 4.7 unrestricted through their native platforms, you would pay between $100 and $200 per each model. That adds up to about $600 of AI that CloudWise Academy is allowing students to use. Inside the class, you have all three. You pick the model you want to work with and that is it.

The framing matters for anyone who has already paid for one AI subscription and assumed that was the full picture. The flagship models from OpenAI, Google, and Anthropic are each sold separately and priced at a premium tier. Erwin consolidates that access so students are not managing three separate premium plans to work at the frontier of what current AI can do.

Getting into Erwin is a short process. Go to erwin.cloudwiseacademy.com. Click the sign up button. Enter your name, username, your email, and a password. Then hit continue. It sets up your account and asks you to log in. That is it. You are in.

Once you are logged in, you are looking at an interface that works like any AI chat you have used. The model selector is where you choose between GPT 5.4, Google Gemini 3.1 Pro, and Claude Opus 4.7. Start chatting with the model you selected and you are working with the most capable AI models on the planet from a single login.

The speaker closes with a clear two-part action: open ChatGPT and try the ImageGen2 workflow shown in this lesson, then head to erwin.cloudwiseacademy.com and sign up to access Erwin. Both tools are available right now. ImageGen2 for marketing visuals that look like they came from a top studio. Erwin to put every leading AI in your hands.