08:14Karolis · Tutorial
The Claude Skills That Run My Entire Content Workflow
A creator gives away the five Claude Skills he built for thumbnails, content research, carousels, motion graphics, and scripts.
July 1stThe argument in one line.
The real leverage in AI-generated UGC is using Claude to batch-multiply prompts so a single MaxFusion session produces 50 unique posts across multiple niches in under an hour.
Read if. Skip if.
READ IF YOU ARE…
- You sell a physical product, SaaS, ebook, or membership and want organic social content without hiring a content team.
- You already know GPT Image 2 exists but have been generating images one at a time instead of building a reusable flow.
- You want to understand how the face-covering phone trick makes AI-generated UGC pass as authentic Instagram content.
- You are considering reselling AI content creation as a service and want a concrete workflow to demo to clients.
SKIP IF…
- You need long-form video content — this is specifically a slideshow-post and short-form video workflow.
- You are not willing to pay for MaxFusion Pro or use GPT Image 2 API credits.
- You want a platform-agnostic approach — this workflow is tightly coupled to MaxFusion as the node canvas.
The full version, fast.
The creator uses Claude to write and batch-multiply image prompts, then feeds them into GPT Image 2 via MaxFusion's node canvas to generate a library of AI UGC images in one session. The key organic-look trick is having the AI model cover her face with a phone. Ten model variations combined with multiple hook text angles yields 50+ unique posts from a single canvas session. The same flow adapts to any niche by swapping prompts. A final segment shows how Seedance 2.0 can animate the static images into short-form video clips.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 00:44
01 · Introduction
Revenue proof ($62K/month RevenueCat), niche versatility claim (product, SaaS, ebook, membership, service), like-gate ask.
00:44 – 01:12
02 · The Setup
Tools: Claude + GPT Image 2 inside MaxFusion. Key pro tip: model covering face with phone is the organic-look trick that makes AI-generated content pass as authentic UGC.
01:12 – 05:21
03 · Building the Content Flow
Live demo — Claude writes base prompt, GPT Image 2 generates first image, Claude batches 10 environment variations, food slide prompts generated from Prompt Vault, Hook Library shown, post assembled and published live to TikTok. Volume math: 50+ posts per session.
05:21 – 06:36
04 · Niche Swap
Entire workflow replicated for skincare — base skin close-up prompt, skincare product slides (serum, moisturizer, gua sha, skin foods). MaxFusion Pro plan and 133 free GPT Image 2 generations mentioned.
06:36 – 08:06
05 · Animating the Images
Seedance 2.0 added as video generation node. Mirror selfie animated into 5-second clip. Food images regenerated with motion-friendly lifestyle framing. Final short-form video shown and TikTok result shared.
Lines worth screenshotting.
- Having an AI-generated model cover her face with a phone is the single detail that makes the image look like authentic UGC instead of an AI render.
- Claude's most valuable role in a content pipeline is not generating images but writing and batch-multiplying the prompts that feed image generators.
- 10 environment variations times multiple hook text angles equals 50+ unique posts from one session — the math makes volume effortless.
- A node-flow canvas lets you swap the entire niche by changing a few prompt nodes, not rebuilding from scratch each time.
- Posting the exact same content format to a brand-new TikTok account and getting traction validates the format independent of audience size.
- Seedance 2.0 extends the UGC workflow into short-form video using the identical hook structure — one more output type from the same assets.
- The hook text on the first slide is the only variable that creates a new post; change the hook, keep everything else, and you have a fresh angle.
- Organic UGC drives app installs at scale because it looks like a real person's content, not an ad — the production strategy is to mimic authenticity, not quality.
- Prompts that specify film grain, natural shadows, and no props are doing more aesthetic work than any post-processing filter.
- Selling AI content creation as a service is viable because the client-facing deliverable looks like a week of professional content work.
Claude as prompt factory, not image generator.
WHAT TO LEARN
The bottleneck in AI content at volume is not image quality — it is prompt variety, and Claude solves that bottleneck by writing and multiplying prompts on demand.
- Claude's most useful role in an image pipeline is writing image prompts, not producing images — separating these two jobs lets each tool do what it does best.
- Batching 10 environment variations in one Claude request costs the same as generating one and gives you a full model library without repetitive manual prompting.
- The organic-UGC aesthetic is a prompting choice, not a post-processing step — film grain, natural shadows, and no props baked into the base prompt beat any filter.
- Covering the subject's face with a phone in the image prompt is the single detail that makes AI-generated content pass as authentic user-generated content in a feed.
- A node-flow canvas turns prompt-to-image into a repeatable, swappable system rather than a one-off task — the niche changes, the structure does not.
- Hook text is the cheapest content variable: the same set of images with different hook text on slide one is a new post with a new angle for the algorithm.
- Adding a video generation node downstream of the image flow converts the same assets into short-form video clips, extending the session's output without new creative work.
Terms worth knowing.
- UGC (User-Generated Content)
- Content designed to look like it was made by a real consumer rather than a brand or agency. In this context, AI-generated images engineered to mimic the aesthetic of organic social posts.
- MaxFusion
- A browser-based node-flow canvas for AI content production. Users connect prompt nodes to image or video generation nodes and run them in batch.
- Seedance 2.0
- An AI video generation model that takes a static image as a reference and animates it into a short video clip. Also referenced as Kling 2.0 in the MaxFusion interface.
- Node flow
- A visual programming paradigm where each step in a workflow is a discrete node connected by wires, making it easy to swap components without rebuilding.
- Hook text
- The first line of text overlaid on a slideshow post's opening image. Determines whether a viewer swipes through; changing this text alone creates a new post angle.
- Base model prompt
- The master GPT Image 2 prompt that defines the human subject's appearance, outfit, and camera angle. All environment variations derive from this single prompt.
Things they pointed at.
06:35productSeedance 2.0
Lines you could clip.
03:15
“Volume is everything. So what you do is take the base prompt back into Claude and ask it to change the environment.”
03:37
“Week's worth of content done in just one sitting.”
02:10
“The model needs to be covering her face with her phone. This one detail is what makes these images look like they came from a real person's Instagram instead of an AI generator.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
analogystory
If you're running any online business, organic social media content like this is one of the best ways to get customers. It's exactly how these apps are getting thousands of new users every single day. So I built this entire workflow using cloth and GPT image to to generate these posts in minutes.
First, I'll show you how to create viral slideshow posts like these, and then I'll show you how simple it is to swap the whole thing for a completely different niche. Whether you're selling a physical product, software, an ebook, membership, this works for all of it.
And you can even sell this as a service, which is what I'm doing myself besides running this YouTube channel. I spent a good amount of time designing the prompts and building this whole workflow, so all I ask in return is for you to drop a like on this video. For everything today, I'll use Claude and GPT image two.
Now you can use GPT image two directly on Charge GPT, but I run the whole workflow inside Max Fusion because of this node flow setup right here, which makes it super easy to swap things out for your own product or niche. And also, you can animate your images into videos with one prompt. I leave this entire flow that I built together with the link for Max Fusion in the description below.
But first, make sure you watch this whole video so you understand how to do it. First thing we need is our base model prompt. This is what tells GPT image two exactly what kind of person we want to generate, and it needs to look organic and real.
Things like slight film grain on the image, natural shadows, nothing too polished. But the most important thing, and this is the pro tip that makes the biggest difference, the model needs to be covering her face with her phone, like in this image right here. This one detail is what makes these images look like they came from a real person's Instagram instead of an AI generator.
So I'll go to Claude and write this prompt to generate the full base prompt. And now we can generate our first image in GPT image two. So now let's just jump into Max Fusion and build the actual flow.
Go to new project, select flows at the top, and create a new flow. Click the plus icon below, select prompt, and then plus icon again. Select image generator and connect them together.
Make sure to select GPT image two nine by 16. Set the quality to high. Drop in the prompt and hit generate.
And there we go. Our first base image is done. Now for this content strategy, volume is everything.
So what you do is take the base prompt back into cloth and ask it to change the environment, and I get a new prompt with a different scene. I'll just drop it back into the flow, and we have another image.
Now let's ask cloth for 10 different environments at once. Drop all of these prompts back into GPT image two.
Let it run. And now you have 10 plus images with completely different environments all done in under ten minutes. Save them all to your computer, and that's your model library ready.
Now for other slides, in this case, we need food images. Since this first example is in the fitness niche where I'm gonna be optimizing this post to promote a weight loss app. Here are all of the prompts I'm using to generate these meal images, And I got all of these prompts just using Cloth.
Whatever type of image you need, just describe it, and Cloth handles the prompting for you. So I'll click the plus icon, select prompt, plus icon again, select image generator, and repeat that for each food prompt.
Now here's where the volume comes in. I've got 10 model shots with different environments and all the food slides done. Every time I change the hook text on that first slide, that's a brand new post with a completely different angle.
So when you add it all up, 10 model variations, multiple hook angles, you're easily looking at 50 plus unique posts from a single Canvas session. Week's worth of content done in just one sitting.
And if you want even more variety, you can just ask Claude to generate new set of food prompts that you can use on GPT image too. So now that we have all the images ready, I'm gonna go and post this on TikTok and show you the results by the end of this video. And, again, all of the prompts together with this whole flow that I'm building right now is in the description below.
Inside, you'll find the base model prompts, different environment prompts, all the food slide prompts, and the full Max Fusion Canvas you can import directly into your account.
You just swap the prompts for your own product and you're generating today. I also included some of the high performing hook ideas that you can use straight away because the hook is what makes or breaks whether anyone swipes through your post. So I'll just grab my phone now and open TikTok right now, and I'm gonna use one of the hooks from the list.
I'll just select all of my images in order from the first intro shot to the last statistics shot, introduces the weight loss app. I'll add the text overlays directly on TikTok using the basic TikTok font, which honestly looks the best. Always make sure you pick some sort of audio when posting a slideshow.
I'll write a simple caption, something like what I eat in a day, add hashtags, and just hit post.
And here's what the finished slideshow looks like. This is the exact format that is getting hundreds of thousands of views. And as you can see, creating content like this is very simple.
Now since everything lives in one flow, you can swap the food prompts to different meals, change the background, change the outfit, and suddenly have hundreds of different image combinations. Just add different hooks and different text on top, and you literally have a content machine for your business. Now here's how simple it is to change this for a completely different niche.
Let's say you're selling a skincare product. So instead of the mirror selfie for the first slide, we're going with a close-up shot of glowy healthy skin. This format is massive in skincare, and it stops the scroll immediately.
Here's the prompt I'm using to generate the base image. So, again, let's just add a new prompt node, connect a new image generation, make sure GPT image two is selected, drop the prompt in, and hit generate.
Now instead of meal slides, we're generating skin care shots. Here are all of the prompts I'm using for that, and here are all of the new images we just made. And if you're not happy with how the first generation came out, just hit generate again, and you'll get a completely different result in minutes.
With Max Fusion Pro plan, you'll also get a 133 free GPT image two generations, and you recently lowered the cost per generation across the board as well, so you're getting significantly more for the same price. Same notes, same setup, just different prompts. This is why I love using the flow setup on Max Fusion.
Swapping everything for a different niche is super simple, and you have the whole thing in one place. Again, you can generate multiple different variations for the intro and for the rest of the slides. Add the text on top when posting, and that's it.
Now here's the best part. You can actually animate these images and turn them into full videos using CDNs two point o. So on top of the slideshows, you can also create short form video content using the exact same hook and structure.
I'm going to animate this mirror selfie and show you how simple it is. Add a new prompt note and a video generation note.
Connect them together. Select Sealands two point o, five seconds, nine by 16. And for generations, select references.
Then select the generated image you want to animate and connect it to the video generation note as reference. And here's the animation prompt I'm going to use. And in about five minutes, we get the animated version.
Now for the food images, since this is a video, we want them to feel more lifestyle, less static. So instead of flat overhead shots, we're generating images with the food being held in hands, different surfaces, different angles.
And here are the prompts I'm using for these. And we have our new set of images done. Now I just saved this clip into my computer, drop it into my editing software, add the food images, put the text overlays on top, and we have a full short form video ready to post.
And the slideshow that I posted on TikTok while recording this is actually performing pretty well for a brand new account. This whole workflow I just created is free and linked in the description below. If you want to see a more in-depth tutorial on generating AI UGC videos using Clawd and Cdance two point o, you can just click here to watch it right away.
Thank you so much for watching, and I'll see you in the next
The Hook
The bait, then the rug-pull.
A revenue dashboard showing $62,288 in monthly app subscriptions. A TikTok slide with a woman holding a phone in front of her face and the caption "I could never eat like you." The creator opens by connecting these two things — and what follows is a complete walkthrough of the AI content engine behind both.
Named ideas worth stealing.
01:12model
Base Prompt to Environment Batch to Hook Swap
- Write one base model prompt in Claude (appearance, outfit, camera angle)
- Ask Claude to generate 10 environment variations of the same prompt
- Generate all images in MaxFusion in one run
- Generate niche-specific content slides (food, skincare, etc.) separately
- Swap hook text on slide 1 to multiply post angles
The three-layer system that turns one Claude session into 50+ unique posts.
Steal forAny product category where UGC social proof drives organic discovery
01:40list
Organic Realism Checklist
- Slight film grain
- Natural shadows
- Nothing too polished
- Model covering face with phone (most important)
- Shot-on-iPhone aesthetic in prompt
Prompt parameters that make GPT Image 2 output read as authentic UGC rather than AI-generated.
Steal forAny AI image prompt for social-first content
How they asked for the click.
VERBAL ASK
07:38link
“This whole workflow I just created is free and linked in the description below.”
Triple CTA: description link to Prompt Vault (creatorkarro.com/vault), MaxFusion affiliate link, and an end-card click to a related AI UGC video. The free Prompt Vault is the lead magnet.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
AFFILIATECommission earned if you click.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
