--- title: 'I Built an AI Content Factory in Claude That Runs My Social Media' source: 'https://youtube.com/watch?v=-QCrU_fmMT4' video_id: '-QCrU_fmMT4' date: 2026-08-04 duration_sec: 708 --- # I Built an AI Content Factory in Claude That Runs My Social Media > Source: [I Built an AI Content Factory in Claude That Runs My Social Media](https://youtube.com/watch?v=-QCrU_fmMT4) ## Summary The video demonstrates how to build an AI-powered content factory using Claude, automating social media management tasks such as writing Telegram posts, creating Instagram carousels, and generating prompts for image generation. The creator shares practical workflows, tools, and prompts to produce human-like content efficiently. ### Key Points - **AI Content Factory Overview** [00:02] — The creator replaced a $150/month copywriter with Claude ($20/month) to write Telegram posts, and uses AI to handle various content tasks. - **Setting Up Claude Projects** [00:44] — Create a project in Claude to store context about your brand, audience, and examples, so the AI writes consistently without re-entering info. - **Humanizing AI Text** [01:13] — Use a special prompt in Claude's instructions to remove obvious AI signs like uniform sentence length and typical AI phrases, making text sound human. - **Providing Context Files** [01:42] — Upload a file with project description, target audience, and post examples to help Claude understand the audience and style. - **Rewriting Posts** [02:08] — Paste a post to rewrite and add the 'rewrite' command; Claude produces content in your tone of voice, allowing a month of content in half an hour. - **Transcribing Videos** [02:21] — Use free tools like TurboScribe (3 free transcriptions/day) or Google Gemini for summaries to repurpose video content into posts. - **Creating Carousels** [03:45] — Carousels are an easy traffic source. Use Claude with specific prompts for image characteristics and composition to generate carousels. - **Finding Prompts** [05:25] — Use prompt aggregator sites like MyGen Gallery to find prompts for AI images, or ask Claude to generate a prompt from a screenshot. - **Using Aggregator Services** [08:08] — Syntax service aggregates multiple AI tools (Nano Banana, Midjourney, etc.) for one subscription, simplifying access. - **Integrating with Design Tools** [08:47] — Connect Claude to Figma or Gamma via connectors to create more polished carousels directly. - **Mass Posting** [11:04] — Use cloud phone services like FGD to manage multiple Instagram accounts (one per phone) to avoid bans. ### Conclusion The video provides a comprehensive workflow for automating social media content creation using Claude and various tools, enabling creators to produce large volumes of human-like content efficiently. ## Transcript designers and entrusted all these tasks to Loud. One project, a few prompts, and the neural network does what was once given to an entire department of people. Let's take a look at how it all works for me in this video. All information in this video financial transactions or other actions shown in the video. The main victory. Now Claude writes posts for me on Telegram channels. If before I hired a copywriter for this and paid him $150, now this task is performed by Claude, who costs $20 a month. And besides this, he takes on a lot of other tasks. And yes, I solved the problem where the text is clearly visible as being written by a neural network. And no, I did n't just get rid of the long [music] dash. We go to the cloud and select projects. We need to create a project so that the neural network remembers data about us and we don’t have to enter it again each time. Click on new project and name it something. Click create project. And then we need to give the neural network a program so that it writes like a human, and not like a neural network. That is, just like showing a child that this was written by a neural network, and this is what this person wrote. You don't need to write like a neural network , you need to write like a human. To do this, click on the instructions and then add this prombe. Added and click save. This is a prom for writing human texts. Specifically, it removes such obvious signs as sentences of the same length, phrases that are clearly created by neural networks, and so on. And then we add this file to the files section. And it contains three parameters. This is a description of our project, a description of the target audience, and examples of posts. This way, the neural network will understand what audience we are working for and will see references. And, accordingly, it will be easier for him to understand what to write to our audience. Now we can give the neural network a post to rewrite. She will write the original version with our t of voice. For example, I'll copy from this channel, paste it here, and add the reite command at the very end. I choose the sonet version. And so, in literally half an hour, we can fill our Telegram channel with content for a month in advance. Copy and paste into your channel. Or we can take the video, transcribe it and rewrite it in our Ton of voice. This is done as follows. Let's say I found a video that I want to transcribe. Copying. And then I can go to the Turbos Scryp service. It's free. Click on transcribe. Then we select our language Russian. Click on the link icon and paste the link to our video here. The service is free. Three transcriptions are available per day The service is free. Three transcriptions are available per day , which is 9093 per month. Or go to Google Gini, again, add our link and write " make a transcription of the video." Gini won't do a transcription for us because it's against Google's rules. However, he will give us a brief summary. And, in fact, we understand what we are talking about here. And we can give this text to the lod. Or use a bot. Bukvitsa is also a bot. Bukvitsa is also a video transcription software. So I gave him a video from Instagram, and he, accordingly, made me a transcription. Moreover, this bot works with both files and links. Here, for example, I gave him a file, and here is the link. I have collected the transcription tools I use in one post. To do this, you need to go to my open need to go to my open Telegram channel and find this post. And here you see links to all these transcribers and go and use them. The link to my open Telegram channel is in the video description. Let 's move on to creating carousels, which is the easiest source of traffic in 2026. They generate no less traffic than the others, but they are much easier to make. This means we've acquired accounts and can now create carousels en masse to drive traffic to our channels. To do this, we do everything the classic way. Let's create a new project. Here in the instructions we add a special prompt, which I will give a little later. Here too, a special prompt is required, as well as a transcription of the video from which we will make the carousel. First of all, we need to add the prompt from this table here. This is a prompt for the characteristics of the carousel image, that the characteristics of the carousel image, that is, what color, what font, and so on. Because by default Claude makes very mediocre carousels, but with this prompt he makes good ones. And then we copy the second prompt and paste it here. This is already a prompt for the carousel composition, so that when composing the text, the neural network first creates intrigue and provides a denouement only at the end, and then a call to action. This is how it works. The neural network first asks us a few clarifying questions, and then produces a pretty decent carousel, minimalistic, but with the colors we need . And at the very end we get the slides. We can download them and upload them to Instagram. I won't go into too much detail about this promt, because I talked about it in this video. Link in the upper right corner. Click and explore if you are interested. Well, you can find the table with the products I just showed in this post on my free Telegram channel. Click on the link in the description, go ahead and pick it up. Compiling prompts for video display. Yes, if you see some neuroslob and want to generate the same one, but don't know the promt, you can compose it using claude. Or first browse through the websites that aggregate prompts. For example, there is aggregate prompts. For example, there is Mayen Gallery. It's a site like Pinterest, but it also provides a prompt for all of these images. For example, we click on these images. For example, we click on this and see the prompt with which this and see the prompt with which it was made, as well as the neural network in which it was performed. For example, this slop was performed in the chat GPT image using this prompt. And here we can simply search or select a belt and enter the description of the neuroslope we need. For example, I'll enter Pixcar. For those who aren't in the know, the Pixar-style neuroslob was very popular this winter and spring. I promoted my Instagram account so much just by watching Pixar cartoons. And we see what images are here. For example, click on this one . And to create Elon Musk in the Pixar style, we need a Pixar style, we need a prompt like this. And it's made in nanobanana. Or this is also done in nanobana and here is Promptt. Or we can take a screenshot of the neuroslope we're interested in, give it to Claude, and Claude will create a prompt for us. For example, I see a video by Potap FFX, a popular blogger. And here the face remains human, but everything around becomes LEGO-style. We go to the cloud, select a new chat. There is no need to create a new project. We throw our screenshots into the cloud [music] [music] . I made a before and after so that the neural network could understand what the original image was and how it needed to be changed. And then we type in: "I want to make the same Lego effect while keeping my face in my photo." Write me a prom for nanoban. Note that the first and second photos are before and after. The person's posture remains unchanged. Use databases of remains unchanged. Use databases of aggregator sites, prompts like Mayen AI. And aggregator sites, prompts like Mayen AI. And we launch [music] then go to nanobanana, insert [music] this prompt and insert the photo that we want to change. I'll insert my own. We send and receive an image like this. As we can see, Claude coped with the task. By the way, I use nanobana and other services for generating images and videos in the Syntax service. This is an aggregator of all the most popular neural networks. And here you can find Nannobana, neural networks. And here you can find Nannobana, Mid Journey, Cdream, Sora GPT Image. From the video here you can find CL, Veo, Runway, Sens, Luma, KFIEL. By the way, SEN is a Sens, Luma, KFIEL. By the way, SEN is a test a huge number of neural networks without having to purchase a subscription for each of them. The link to Syntax will be in the description. Carousels can be decorated even more beautifully. To do this, it is necessary to create an integration between Claude and, for example, Figma or Gamma. To do this, select the Customize section, then connectors and click on the plus sign. Next, Browse Connectors. And we see services that we can integrate into the LOD. This is Canvas, this is Figma. And you can also find many other programs that we use in everyday life. I need gamma. Here she is. Gamma, I see. And I click plus. Here we should already be registered. Yes. We choose to allow. Allow. And we see that we have added gamma. Now let's return to the chat section. Here you need to make sure that the connected application is activated. Here we see connectors and here is the gamma. And it is important that the slider is turned on. And then, without any projects or complex projects, we'll extract the key extract the key ideas from my ideas from my text and create an Instagram carousel in gamma. And I will take this text, in gamma. And I will take this text, which I transcribed in capital letters. edit it a little and immediately upload it to Instagram. Select [music] Edit in Gamma. The aspect ratio needs to be changed here , but I like the essence, the composition itself. And I especially like how Gamma presents statistics and how it makes infographics. We can also integrate Canvas, Figma, and other services for generating images and carousels. I recommend testing a few to see which ones work best for you and which ones generate the most successful carousels. And we can move on to mass flooding. For this purpose, we choose the FGD cloud phone service. And we have here [music] one phone, one Instagram account. Accordingly, we go to each phone and load our carousels. For those who aren't aware, Instagram is currently cracking down on multiple accounts, and the current limit is one account per phone. And the issue has to be resolved like this. If you're currently wondering to generate massive amounts of social media traffic, then you're interested in content farming. I talked about this in great detail in this video. Click. in this video. Click. Enjoy watching.