Sora AI Videos for TikTok Growth: The VA-Powered Content Engine That Scales

David Attiaas built a system to promote STOPPR—a women's sugar tracking app used by over 235,000 women to quit processed sugar through a 90-day science-based reset program—using Sora AI-generated videos posted by virtual assistants.

The strategy doesn't rely on traditional content creation. It's built on AI content arbitrage: generating high-quality videos with Sora Pro, removing watermarks to bypass platform detection, and deploying virtual assistants to post at scale across warmed-up TikTok and Instagram accounts.

This isn't about going viral once. It's about creating a repeatable system that platforms can't yet detect—while they still can't.

The Quality Arbitrage: Why Sora Pro Beats Sora Free for Scaling TikTok Content

Most creators experimenting with AI video use Sora Free. David took the opposite approach: pay $200/month for ChatGPT Pro to access Sora Pro.

Sora Free works—but inconsistently. The quality varies dramatically between generations. Some videos look professional, others have audio issues or visual artifacts that immediately signal AI generation.

Sora Pro solves the consistency problem. From David's experience: 95% of videos generated by Sora Pro meet quality standards for organic TikTok posting, compared to less than half with Sora Free.

When you're scaling to 5 videos per day across multiple accounts, quality consistency matters more than initial cost.

One unusable video breaks your posting schedule. 5 broken videos in a week destroy momentum.

You can access Sora Pro by subscribing to ChatGPT Pro for $200/month. You'll also need a VPN with US IP address access (David uses Surfshark) since Sora availability varies by region.

Bypassing TikTok AI Detection: The Watermark Removal Strategy

Here's the critical detail most creators miss: Sora videos come with watermarks embedded in every frame. These watermarks don't just identify the content as AI-generated to viewers—they potentially signal to TikTok's detection systems that the content isn't human-created.

TikTok has an advanced option to flag content as AI-generated. David tested both approaches:

• Videos flagged as AI: Lower reach

• Videos posted without the flag: Better performance

But you can't just skip the flag and post Sora videos directly. The watermark on every frame is the tell.

The watermark isn't added once—it's embedded in each individual frame of the video.

David's solution: watermark removal before posting.

After testing multiple paid and free watermark removal tools, he found Vmake.ai to be the most effective, despite not being free. As a paid Pro subscriber himself, he considers it worth the investment.

The process: upload your Sora-generated video to Vmake's Watermark Removal Tool, and it extracts each frame, removes the watermark, then concatenates the frames back into a clean video. Processing takes 5-10 minutes per video.

Alternative tools David tested but found less effective: Magic Eraser and other free options. For scaling content, consistency in watermark removal quality matters.

Once the watermark is removed, TikTok can't detect the content as AI-generated. The detection gap remains open—but it requires this preprocessing step.

The Human Signal Strategy: Making AI Content Look Organic

Generic AI video workflow: Generate video → Remove watermark → Upload directly → Hope for reach.

David's workflow adds deliberate human signals:

1. Generate video with Sora Pro (no text on screen in the prompt)

2. Remove watermarks with Vmake AI

3. Enhance audio with Adobe Podcast if needed

4. Add text overlays directly in TikTok/Instagram, not in external editors (VERY IMPORTANT)

5. Post from warmed-up accounts (purchased, not self-created)

The critical detail: text overlays added within TikTok itself, not CapCut or other video editors.

Why? Platform algorithms reward native editing. When you edit within the app, you're signaling real user behavior—not bot activity.

Every friction point removed, every shortcut taken, makes content look less human. David adds friction back intentionally. The goal isn't efficiency—it's algorithmic trust.

For audio enhancement, if Sora's voiceover sounds too robotic, David's VAs use Adobe Podcast's Enhance Speech tool, which offers a free 30-day trial. One click cleans background noise and makes voices sound more natural.

The Comment Firewall: Preventing Perception Collapse on TikTok

The biggest threat to AI-generated content isn't detection by algorithms—it's detection by humans in the comments.

One commenter saying "this is AI slop" creates social proof that the content is fake. More users pile on. Engagement shifts from positive to negative. The video dies.

David's solution: preemptive keyword blocking on TikTok.

On TikTok, you can filter specific keywords before comments go live (Settings → Privacy → Comments → Filter specific keywords).

His complete blocked word list:

ai, bot, poor, suck, fake, cringe, weird, ugly, cheap, low quality, trash, scam, bot generated, artificial, lame, mid, low effort, hate, garbage, dumb, stupid, not real, unrealistic, awful, disgusting, broken, looks off, so fake, ai video, clearly ai, not human, deepfake, slop

The strategy isn't hiding AI usage—it's preventing negative social proof from accumulating. Users who would engage positively never see the negative comments. The perception remains intact.

Most creators think of comment moderation as reactive. David treats it as infrastructure—set up before the first post, not after the first attack.

The Virtual Assistant Multiplier: From Creator to Operator

Creating 5 videos per day is manageable. Editing them, timing posts for US peak hours, monitoring performance, and adjusting strategy across multiple accounts isn't.

David doesn't create content. He generates videos with Sora and delegates everything else to virtual assistants in the Philippines.

The delegation framework:

• Video generation: David handles this (requires ChatGPT Pro access, VPN, prompt engineering)

• Video editing: VAs remove watermarks with Vmake AI, enhance audio with Adobe Podcast, add app footage

• Platform posting: VAs add text overlays in-app, post at scheduled times

• Performance monitoring: VAs flag underperforming videos (<100 views after 1 hour)

The timezone difference becomes an advantage. While David is asleep in France, his VAs in the Philippines post during US peak hours (7-10 PM EST). They post at 7, 8, 9, and 10 AM Philippines time, which translates to 7, 8, 9, and 10 PM New York time.

The critical instruction: post 1 video per hour, up to 5 videos per day. If any video gets stuck below 100 views after 1 hour, stop posting and report back to David so he can take a look and warm up again if needed.

This isn't just efficiency—it's risk management. One shadowbanned video could signal platform distrust. Stopping immediately prevents account damage.

The App Plug Formula: Transformation + Proof

Every video follows the same structure:

• 10-15 seconds: Sora AI-generated lifestyle content (what I eat in a day, transformation footage)

• 5 seconds: Screen recording of the STOPPR app

Some human examples of Stoppr footage can be viewed here by Eve Olivia, Paullina Ruban, and Orinda.

The hook isn't the app. The hook is the lifestyle content that makes viewers stop scrolling. The app plug is the conversion mechanism at the end.

David provides pre-made app recordings in a Google Drive folder. His VAs don't create new footage—they splice existing recordings onto the end of Sora videos.

Example prompt that generated a successful video: "Fast paced vlog in Bali video of gen z woman sharing her 7 days no processed sugar journey. No text on screen."

The video shows a woman eating healthy food, exercising, enjoying vacation. At the end: a screen recording scanning a cheesecake with the STOPPR app.

The narrative arc: aspiration (lifestyle content) → solution (app that helps achieve it).

STOPPR helps women reset their cravings, rebalance dopamine, and take back energy, clarity, and control through structured mindful eating challenges, dopamine-balancing exercises, guided meditations, and real-time progress tracking. The lifestyle content shows the transformation; the app shows the tool that enables it.

The Grammar Mistake Hack: Imperfection as Authenticity Signal

One detail buried in David's VA instructions: intentional grammar mistakes sometimes increase views.

Perfect grammar signals professional content or bot activity. Minor errors signal real human posting—especially in Gen Z-targeted content.

The strategy isn't making every post have errors. It's allowing imperfection when appropriate. Not correcting "ur" to "your" or "alot" to "a lot" in overlay text.

Most brands obsess over polish. David recognized that excessive polish triggers skepticism. Slight imperfection increases perceived authenticity.

The Account Sourcing Reality: Why Self-Warming Fails

David doesn't warm up TikTok and Instagram accounts himself. He buys them from someone on Twitter/X.

The logic: account warming requires posting regularly from a consistent location over weeks. From France, he'd need either:

• US-based virtual assistants to create and warm accounts

• Complex VPN/eSIM setups to simulate US user behavior

• Months of preparation before launching content strategy

Buying pre-warmed accounts removes the bottleneck entirely. The cost is negligible ($20 per account) compared to the opportunity cost of delayed execution.

This is the difference between operator thinking and creator thinking. Creators warm their own accounts because it feels more authentic. Operators buy warmed accounts because authentication already happened—they're just purchasing the result.

The Timing Window: Why This Works Now But Won't Forever

David's entire strategy relies on watermark removal closing the detection gap. Once you remove the Sora watermark from each frame, platforms can't identify the content as AI-generated.

But this advantage is temporary. As watermark removal becomes more expensive or platforms develop frame-by-frame analysis that doesn't rely on watermarks, the window narrows.

His acknowledgment: "Maybe TikTok can't detect yet those Sora videos are AI generated."

The word "yet" is doing heavy lifting. He knows this is temporary arbitrage. The strategy isn't build a sustainable brand—it's extract maximum value during the detection gap.

When the tools become prohibitively expensive and every video is an AI video, the playbook shifts. But the underlying principle remains: find what platforms can't detect, scale before they can, move to the next gap.

Personally, I don't think any social media can ever detect AI-generated videos and open-source Chinese models might close the gap so you can learn this skill once and scale to the moon.

The Documentation Principle: Systems Over Heroics

David's entire operation runs from a single Google Doc titled "Onboarding."

Inside: links to tools, step-by-step workflows, keyword block lists, posting schedules, editing instructions, example text overlays.

Everything is documented before the first VA is hired. The system works because it doesn't depend on his daily involvement.

Most founders scale by working more hours. David scaled by making his process transferable. The Google Doc is the product as much as the videos are.

Creating one more video has linear returns. Creating one more system has exponential returns.

Sora 2 Arbitrage Playbook: The Replicable Pattern

David's strategy works for STOPPR because it follows a proven formula:

1. Find a content quality gap (Sora Pro vs. Sora Free or UGC videos)

2. Exploit a platform detection gap (Remove watermarks with Vmake AI)

3. Add human signals intentionally (in-app editing, minor imperfections)

4. Delegate execution systematically (VAs handle everything repeatable)

5. Monitor and adjust quickly (stop posting if engagement drops)

The specific tools will change. Sora will evolve. Different watermark removal tools will emerge. Different apps will need promotion.

But the pattern stays the same: find temporary advantages in AI tools and platform limitations, then scale aggressively before the window closes.

This isn't traditional marketing. It's not about brand building or audience relationships or authentic connection.

It's about recognizing that platforms and AI tools move at different speeds—and capturing the value in the gap between them.

For now, that gap is wide open. David's building his business inside it.

The question isn't whether the gap will close. It's whether you'll move fast enough to benefit before it does.

Top Tweets of the day

1/

PLG (Product-Led Growth) for the win.

2/

This comment has very few likes but the comment below it has 1.7K+ likes. Except that one is wrong. Nowadays it is easy to go viral by saying incorrect stuff.

3/

Advice straight from the source. This guy is a researcher at OpenAI behind Sora.

Rabbit Holes

1. Easy trick I use to spot AI comments

2. The OpenAI o4-mini Model Interview

3. Sora 2 Prompting Guide

What’d ya think of today’s newsletter? Hit ‘reply’ and let me know.

Do me a favor and share it in your company's Slack #marketing channel.

First time? Subscribe.

Follow me on X.

More Startup Spells 🪄

1. LinkedIn Growth Hacks: 145,000 Followers in 14 Months (LINK)

2. TablePlus's Smart Retort To Pirates (LINK)

3. Windsurf's Go-To-Market (GTM) Strategy: From VSCode Fork to $3B Rumoured Exit to OpenAI (LINK)

4. How Google stole OpenAI's spotlight (LINK)

5. Jacky Chou's SEO Survival Playbook: Manipulating Google's Algorithm for a Living (LINK)