11:35Jonathan Nilsen · Tutorial
How to Go Viral on TikTok With Your Books
A five-rule breakdown of why book videos go viral, then a live ChatGPT tutorial for turning a manuscript into ten scripts in about 30 minutes.
September 9thThe argument in one line.
A single AI agent can compress six-person agency work — research, design, branding, video, copywriting, and web build — into one afternoon chat, with economics that return roughly 6x on the spend before the first ad dollar is placed.
Read if. Skip if.
READ IF YOU ARE…
- You have a product idea you have never acted on because the startup cost felt too high.
- You are already selling online but paying freelancers for content, design, or web work that feels slower and more expensive than it should be.
- You want to validate demand for a physical product before committing to inventory.
- You are curious how AI agent orchestration works in practice on a real product launch, not just in a demo environment.
SKIP IF…
- You are looking for a platform-agnostic tutorial — this is specifically a walkthrough of Higgsfield Supercomputer and its proprietary models.
- You need a revenue path that does not involve physical product sourcing or social media distribution.
The full version, fast.
Higgsfield Supercomputer is an AI agent that orchestrates multiple specialized models inside a single chat. The host used it to scan real marketplace data and pick lipstick as a high-margin niche, mine thousands of reviews to find the core complaint (dryness), generate a full product spec, produce five content formats engineered to match what is currently going viral in the category, build and deploy a mobile-first waitlist page with AI-generated visuals, and set up the full funnel — all for $121. The result: 833 waitlist signups and 500K Instagram views in three days on a brand-new account, before spending a dollar on ads.
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 – 00:39
01 · Cold open — results first
Social proof before explanation: views, signups, zero paid ads on a 3-day-old account. Sets the stakes before the tutorial begins.
00:40 – 02:19
02 · Stage 1 — Finding the niche
Supercomputer scans Amazon/Etsy/Shopify/Meta Ads, returns three finalists, recommends lipstick. Then mines Sephora/Amazon reviews to surface the core complaint (dryness) and build the product angle around solving it.
02:20 – 05:37
03 · Stage 2 — Content factory
Five content formats identified from real platform data (ASMR, makeup application, GRWM, unboxing, talking gossip), then generated by the Higgsfield orchestrator using Sol and CDance 2.0.
05:38 – 07:58
04 · Stage 3 — The website
Waitlist page (not a store) built and deployed in the same chat. All visuals generated by Supercomputer — no stock photos, no extra tools. $10 custom domain optional.
07:59 – 08:56
05 · Results check
500K views on a 3-day-old account, 833 waitlist signups. Analytics fed back to Supercomputer to double down on ASMR.
08:57 – 10:07
06 · Cost comparison
Freelancer alternative: $3K+ and 3-4 weeks. This build: $121 and one afternoon. Six skill sets, six people on a payroll — vs. one AI agent in one sitting.
10:08 – 11:35
07 · ROI math
833 signups x 10% x $16/unit - $3/unit COGS - shipping/duty/fees/packaging/domain - $121 Supercomputer spend = ~$738 net profit. ~6x return from first wave alone.
11:36 – 12:52
08 · Takeaway + CTA
The point is not lipstick. The cost of trying just dropped to nothing. 8K likes = live business coaching in next video.
Lines worth screenshotting.
- AI agents do not just generate content — they analyze what is already performing before generating anything, which is the difference between content that looks good and content engineered to perform.
- Validate demand with a waitlist before spending on inventory: 833 email signups for a product that does not exist yet is proof the market wants it.
- Six professional skill sets — research, design, branding, video, copywriting, web build — cost $3,000+ and three weeks from freelancers; one AI agent chat replaces all of them for $121.
- The 10% waitlist conversion assumption is a useful conservative floor for modeling whether a product launch pencils out before you commit to sourcing.
- Review mining at scale is the fastest way to find a product angle: when the same complaint appears across thousands of reviews for every brand at every price point, that complaint is your product.
- Sourcing on Alibaba at the higher end of the price range ($3/unit) is the right conservative assumption for a custom product launch — padding the cost model is how you avoid a nasty surprise at fulfillment.
- The feedback loop that used to require a performance marketing team — post, read analytics, double down on what is working — now runs inside one AI chat in a single prompt.
- A 3-day-old Instagram account with zero followers can hit 500K views without paid ads when the content is built from the formats that are already going viral in the category.
- The real barrier to starting a business in 2025 is not money or skills — it is whether you actually press send on the first prompt.
- $121 in AI credits versus $3,000+ in freelancers is not a marginal improvement; it is a structural change in who can afford to test a product idea.
Validate the idea before you fund it.
WHAT TO LEARN
The most expensive mistake in product businesses is building before anyone proves they want the thing — this video maps a complete low-cost path around that mistake.
- Start with the complaint, not the product: mining thousands of reviews for the most repeated frustration gives you a product angle that is pre-validated by real buyer language.
- Content format selection is a research problem, not a creative one — the formats most likely to perform are already visible in platform data before you spend anything on production.
- A waitlist page is not a placeholder — it is the test. Capturing emails before you have inventory tells you whether demand is real without the risk of holding product nobody wants.
- The feedback loop is the moat: feeding early analytics back into your content strategy is what separates a one-hit video from a growing account, and it takes one prompt to run.
- The 10% waitlist-to-customer conversion rate is a useful conservative floor for modeling whether a product launch pencils out before you commit to sourcing.
- Sourcing at the high end of the quoted price range is the right default for a first custom order — surprises in fulfillment almost always add cost, not subtract it.
- Six skill sets that used to require six people now route through a single prompt chain — the unlock is not that any one task got easier, it is that the coordination cost between tasks dropped to zero.
Terms worth knowing.
- Supercomputer
- Higgsfield's AI agent platform that orchestrates multiple specialized models — research, image generation, video generation, web build — inside a single conversational interface.
- Orchestrator
- The routing layer inside Supercomputer that decides which underlying model handles each subtask, analogous to a project manager assigning work to specialists.
- Sol
- A character model used within Higgsfield's pipeline to generate consistent human-looking UGC actors across multiple video formats.
- CDance 2.0
- An animation model inside Higgsfield used to generate realistic motion for AI-generated characters in UGC content.
- UGC (User-Generated Content)
- Short-form social video styled to look like organic creator posts rather than polished brand ads — typically more trusted and more likely to go viral.
- Waitlist funnel
- A launch strategy that captures email subscribers before a product ships, validating demand and building a warm audience to sell to at launch.
- ASMR format
- Video content built around satisfying close-up sounds and visuals (mixing, grinding, textures) that triggers a relaxation response — among the highest scroll-stop formats in beauty.
- Meta Ads Library
- Facebook and Instagram's public database of all currently running ads, used here as a signal for what products brands are actively spending money to promote.
Things they pointed at.
Lines you could clip.
02:11
“That's hours of competitor research compressed into one chat. No human picks a niche this easy in just five minutes.”
08:42
“That is a feedback loop you only get from a full performance marketing team.”
09:43
“The people who get rich in the next five years are not the ones with the best skills. They're the ones who move first while everyone else is still arguing whether AI is real yet.”
10:00
“The barrier to start a business just dropped to zero. The only thing left is whether you actually do it.”
11:36
“The point is not that you should go and start a lipstick brand tomorrow. The point is that the cost of trying something just dropped to almost nothing.”
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.
metaphoranalogy
Everything you're about to see came out of the Splunk chat with zero help from me. A product, a live website, motion design, and UGC videos.
A launch across Instagram and TikTok, and a real wait list for something that didn't exist a week ago. And I didn't write a line of code, didn't fill in a single frame, didn't even pick what to sell.
Just gave it a couple of sentences, and once everything was live, I closed my laptop for three days and came back to real views, real sign ups, zero paid ads on an account that was three days old. Hi.
I'm Adil. Let me show you how you can do all of that too.
The first thing I need is a niche. So I'm telling Supercomputer to scan top performing products across Amazon, Etsy, Shopify, and Meta Ads, and find me the best ones for a content driven business, optimized for high margin and easy distribution.
This is really important. By the way, I'm gonna leave all the prompts in the description so you can easily try them yourself. Now let's see what supercomputer is gonna come up with.
So instead of guessing, it's actually going through real marketplace data and ad libraries right now, looking at what people are buying and what brands are spending money to advertise. So it comebacks with three options and recommends lipstick. High margin visual product built for short form, so I tell it to go with lipstick.
And this is where it stops feeling like every other AI tool I've used. It pulls thousands of reviews from the top lipsticks on Sephora and Amazon, and starts looking for the most common complaint.
So the same thing keeps coming out. They dry your lips out, regardless of the brand or the price. So it's gonna build the whole angle around fixing that.
A lipstick that's half lip balm, Squalane, Shea butter, Ceramides, the stuff you'd find in a fancy lip mask, but in a lipstick.
It picks the name. It even generates the full product spec sheet I could hand, uh, straight to manufacturer.
I mean, that's . No human picks a niche this easy in just five minutes.
Okay. So we've got a product, but a product on its own doesn't sell anything. What sells is the content.
So I asked Supercomputer the next obvious question: What kind of content wins in this niche right now?
It pulls the top performing beauty content from the last thirty days, breaks down the hooks, the pacing, the visual styles, and comes back with five formats that are working right now.
We've got ASMR, makeup application, get ready with me, unboxing, and the talking gossip format.
Let's make all five of them. First up, ASMR. And here's the thing.
Supercomputer generated the ASMR beauty videos for our lipstick based on the winning patterns it just identified. Let's see what it gave us.
Look at this. This is the kind of content you literally never scroll past. The grinding, the mixing, the way the paste spreads into the slime.
It's so satisfying that it's almost hypnotic, and that's exactly what's going viral on beauty Instagram right now. The supercomputer knew that because it analyzed what's already working before it generated anything.
These videos easily get thousands of views minimum. Okay. So the second video format is a product test, and this is where it gets interesting.
Inside supercomputer, there's a built in engine that runs your whole content factory. Even though the prompt we just used is short and simple, there's a whole complicated process underneath.
So Heathfield's orchestrator chose which model to use for this task, then a specific skill got pulled in to start the workflow. For creating UDC content, it uses Sol for the character and Cdance two point o for the animation.
And the final output looks consistent and realistic even with the simplest prompt.
I mean, look at how clean this shot is. We're showing how this one shade works for your lips, cheeks, and eyes. With lipstick, the video can be as simple as the product.
No complicated setup. No lengthy explanation. The visuals speak for themselves.
This genuinely looks like something a real creator would post. Now let's see how supercomputer is gonna handle the next format.
And it really got this one right. I mean, no words needed. It feels like a real unboxing on your feet.
So we now have all different types of content for the same lipstick, all ready to post. And every single one was generated based on real data and was actually performing on Instagram right now.
That's the difference between content that just looks good and content that's actually engineered to perform. Let's get it out there and check the results.
Now while everything is running on Instagram and TikTok, we still need somewhere for people to buy this thing. So let's build a website.
We want a waitlist page, not a store. The reason is simple. We're validating demand first and only then selling.
So we're not investing our money in inventory until we know people actually want this. So I'm typing, build a waitlist page for our lipstick brand, optimize for mobile conversions, matching the brand style we've built so far, and let's generate.
Notice I didn't pick the colors. I didn't pick the fonts. I didn't write a single line of code.
Supercomputer handled all of it. And here's the part nobody else does.
Site builders give you a template, but you still have to bring your own content. Supercomputer generated every visual on the site in the same chat.
Banner, product renders, lifestyle, imagery, Zero stock photos here.
Zero extra tools. And it's mobile first because every visitor is coming from Instagram and TikTok. And it's already deployed, so supercomputer pushed it to its own URL automatically.
No hosting to set up, no DNS, no extra tools. It's simply live.
Additionally, I did spend $10 to buy a custom domain just so it looks like a real brand instead of a generated link. Totally optional, but $10 to make it feel legit is worth it.
One more piece left, and it's the one that turns all of this from a demo into a real funnel. I'm pasting the website URL into the Instagram bio, and with that, the funnel is live.
Content on Instagram and TikTok drive traffic to the bio link, which sends people to the waitlist page, which captures them onto the email list. That's the whole thing.
Content brings traffic, bio link routes it, waitlist captures the email, and that list is what we sell to when the product ships. No paid ads, no store, no inventory.
Validate demand first, then sell to a warm list. That's how you launch a brand without burning cash. Now, let's look at the actual results of our work, starting with Instagram.
I mean, look at this one. 500,000 views on an account that was three days old with zero followers.
I mean, the algorithm pushed it because the content was good enough to compete with real creators. That is the whole experiment right here.
Now, here's my favorite part. After the first videos went out, I fed the analytics back up to supercomputer, and told it to double down on what's already working.
It went straight into creating more ASMR. That is a feedback loop you only get from a full performance marketing team. Now, the part that actually matters, the wait list.
833 people signed up for a product that did not exist three days ago.
Every one of them came through a funnel that took one prompt to build. That's proof you can build a real business on top of this. Now let's talk about what this would have cost you a year ago.
A designer for the site, a video editor for the content, and at least three UGC creators. You're looking at, what, 3,000 minimum in three to four weeks of waiting, and that's if the freelancers hit their deadline, which they never do.
I just did all of this in one afternoon from one chat for the price of a dinner. And here's what should make you stop and think.
Research, design, branding, video, copywriting, web build.
That's six different skill sets. Usually, six different people on a payroll. But we just did all of that with one AI agent in front of you, in one sitting.
The people who get rich in the next five years are not the ones with the best skills. They're the ones who move first while everyone else is still arguing whether AI is real yet. The barrier to start a business just dropped to zero.
The only thing left is whether you actually do it. Now I know exactly what some of you are thinking. Okay.
But how much did this actually cost? Because none of this matters if supercomputer is more expensive than just hiring people to do all of it.
Total spend to build everything you just watched, I mean, the product, the brand, the website, the videos, the launch, was a 121 in supercomputer credits.
And to be real with you, that includes the failed generations along the way. Now a $121 on its own might sound like a big number.
So let's actually do the math on what that build is. So we have 833 people on the waitlist.
If just 10% of them actually buy when the product launches, which is a standard conversion rate for a waitlist, that's 83 customers. Each tube sells for $16, so that's about $1,328 in revenue.
To make each lipstick, we sourced suppliers on Alibaba, and we deliberately picked the higher end of the price range, around $3 per unit, because custom orders usually cost more than the base price.
Add shipping, duty, Stripe and ecommerce platform fees, packaging, plus the domain we bought for the brand.
That's another $220. And let's also add back the $1.21 we spent on supercomputer itself, so every single cost is accounted for. That's around $738 in net profit from one twenty that went in.
That's roughly six times your money back, and that's only from the first wave of waitlist sign ups. The views on social media are still growing, and the waitlist is still filling up every hour. So this is the floor, not the ceiling.
Here's what I actually want you to take from this video. The point is not that you should go and start a lipstick brand tomorrow. The point is that the cost of trying something just dropped to almost nothing.
A year ago, building what you just watched would have taken a team, a real budget, and a month of work. Today, it took one chat and $121. If you've ever had a product idea sitting in your nose app and never done anything with it, this is the moment to actually try.
And if you've already got a product or a business, but you're stuck with no views, no sales, and no traction really, I wanna help you fix that. Drop a comment with your name, your age, when you're selling, and what's actually stopping you from growing.
If this video hits 8,000 likes, I'll pick one of you, and in the next video, we'll grow your business live using Supercomputer, from wherever you're stuck all the way to actual results. Link to the same workflow we used and all the prompts are in the description.
And as always, if you found this video helpful, hit the like button, subscribe, and I'll see you guys in the next one.
The Hook
The bait, then the rug-pull.
A single chat. No code written, no frame filled in, no niche chosen. The host closed his laptop after setup and came back three days later to 500,000 views on a zero-follower account and 833 real email signups for a product that still did not physically exist.
Named ideas worth stealing.
00:40list
The 3-Stage AI Launch Loop
- Stage 1: Niche + product spec (marketplace data + review mining)
- Stage 2: Content factory (platform-validated formats, AI-generated)
- Stage 3: Waitlist funnel (validate demand before inventory)
Three sequential phases run through Supercomputer to go from zero to deployed brand in one afternoon.
Steal forAny physical or digital product launch where you want to test demand before committing budget
05:58model
The Validate-First Funnel
- Content (Instagram/TikTok)
- Bio link
- Waitlist page
- Email list
- Launch to warm audience
No paid ads, no inventory, no store. Build the list first; sell when the product exists.
Steal forPre-launch campaigns for any product with a 4-12 week production lead time
10:08model
The 6x ROI Model
833 waitlist x 10% conversion x $16 price - $3/unit COGS - shipping/fees/domain - $121 AI spend = ~$738 net. Presented as floor, not ceiling.
Steal forAny video framing AI investment as low-risk with concrete math
How they asked for the click.
VERBAL ASK
12:27engagement-gated next-video
“If this video hits 8,000 likes, I'll pick one of you, and in the next video we'll grow your business live using Supercomputer.”
Strong. Ties subscriber action to a concrete community prize (live coaching session). Prompts comments with a specific format (name/age/product/what's stopping you), generating both engagement signal and social proof.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
