Creator strategy · OpenArt case study

How OpenArt gets creators to sell one product six ways

Each format gets its own viewer, its own ask, and its own amount of interface on screen. A feature name was seeded across 34 posts in 57 days. The accounts that copied the script got under 500 views each. Everything here comes from public posts.

OpenArt
Creator strategy · 250 public posts · 227 creators

OpenArt handed creators a word. The accounts that said it most faithfully got under 500 views.

“Vibe directing” is OpenArt’s own name for its Director mode. Big creators explained it, a partner credited it, and a cluster of dedicated accounts recited it on the same day.

Viral Anatomy · Case study
Shark Numbers' OpenArt workspace: the same talking-head clip rendered in three backgrounds
Shark Numbers1.9M subscribers1.4M views57.5K
Shark Numbers · one clip, three backgrounds, his own workspace and credit balance in frame

On June 22 a YouTube tutorial with 953 views used the phrase “vibe directing” for OpenArt’s new Director mode. By August 17, 34 posts in our OpenArt sample had said it out loud. Five of them were TikToks that landed on that single day, from accounts that post about almost nothing else.

The word was written by the brand, handed to creators, and repeated on a schedule. That is the most visible thing OpenArt does with creators. It is not the most effective.

This piece is about how one product became six different videos, how the word got planted, which tier of creators actually produced the hits, and why the accounts that recited the word most faithfully got the least for it. By the end you can write the brief for the format that won, and you know which tier to set up first.

34 posts that say “vibe directing”, June 22 to August 17 dot size = views (Instagram hides views) · one row per platform · five TikToks on August 17 YOUTUBE TIKTOK INSTAGRAM Jul 1 Jul 15 Aug 1 Aug 15 Jun 22 · first use, 953 views Jul 3 · Kate Telek defines it Jul 22 · Zadonsky credits OpenArt (partner) Jul 13 · Tech Vorld 98K, 1.0× Aug 4 · ksdealz 77.5K Aug 17 · five copies, 171 to 476 views VIRALANATOMY.COM
Every post in the sample whose transcript contains “vibe directing.” Dot size is views; Instagram hides view counts, so those dots are drawn at minimum size. The first wave is scattered explanations by larger accounts. The last wave is a single day.

What OpenArt sells, and whether this applies to you

OpenArt (openart.ai) is an AI creative suite: image and video generation, music, a Director mode that turns a chat into a storyboard and a timeline, a VFX tool that repaints the background of footage you already shot, and a Characters feature that keeps one face consistent across scenes. The buyer is a solo creator who needs to produce content, or wants to sell content.

Three questions decide whether OpenArt’s playbook is your playbook:

Can the result be filmed?

Yes, and it is spectacular.A talking-head clip becomes a neon lab. A photo becomes a five-minute film. That is why so many of these videos open on the finished output.

What does it save the user?

Skill and production cost.So the scripts sell “you don’t need to know how” and “the output can be sold.”

The category’s known failure?

Character drift and the “AI look.”The strongest reviews open on that failure and fix it on screen.

If your product’s output is something people would watch on its own, the rest of this piece applies. If it isn’t, only the last section does.

relationshipwhat counts as evidencecreatorspostsmedian views
AFFILIATEtracking link or discount code in the description647311,778
PAID#OpenArtPartner, “sponsored by”, or a platform paid label131433,652
PROGRAMCreative Partner Program credit or #openartcpp1919493
CONTESTan entry to one of OpenArt’s own film competitions910196
UNKNOWNno public evidence either wayUnknown is not organic. It means we found no receipt.121133390
Plus one self-described ambassador with one post. By platform: 153 YouTube posts, 57 TikTok, 40 Instagram. YouTube holds 89% of all views in the sample. Medians count posts with a visible view count.

One product, six different videos

OpenArt does not have one creator brief. It has six, split by what the viewer wants when they press play. Each one gets a different first frame, a different amount of interface on screen, and a different ask.

Look at the first frames. The tutorial opens on “AI VFX” and the OpenArt logo. The contest film opens on a pencil sketch of an old man and the caption “He used to be the funniest man I know.” The faceless-channel course opens on a counter reading “+8,361,721 views” next to “$+6,839.” Nothing about these three suggests the same product. That is the design: each line gets a different viewer, a different promise, and a different ask, and no creator is asked to do two of them in one video.

How the word got planted

The first wave is explanation. On July 3, Kate Telek, a smartphone-videography account with 649K followers and no disclosed relationship, opens with the definition.

UNVERIFIEDKate Telek in a cozy room, announcing a new feature
Jul 3 · @kate.telek · TikTok
“there’s a new category in AI video called vibe directing, think vibe coding, but for videos.”0:00
1,568views
649Kfollowers
0:52runtime
UNVERIFIEDProfile and recent posts checked; no commercial tie found. Open the evidence →

On July 22, Andrew Zadonsky, a creative director with a disclosed #openartpartner tag, credits the source on camera. His post is the anchor of the run, and its five steps are the cleanest version of the script.

See it run, once · @zadonsky00 for OpenArt · Instagram · disclosed partner · views hidden by the platform
0:00
Result first
“How did I vibe directed this?”
The term is used as if the audience already knows it. The fastest way to make a word feel established.
→
0:07
Credit the source
“that’s why OpenArt calls it vibe directing”
The brand’s vocabulary is handed over explicitly. That is what makes this seeding rather than slang.
→
0:11
Redefine the input
“I just thought to director, describe my idea, upload a song”
The pitch is the absence of prompt-writing, positioned against every other AI tool.
→
0:22
Voice, not typing
“instead of typing, I used my voice”
A single concrete detail carries the whole “directing, not prompting” claim.
→
0:37
Edit in place
“It moved me on to the timeline, so I could edit it right inside director”
Staying inside one tool is the payoff: the reason to switch, shown rather than argued.

Between them, on July 15, filmmaker Josh Glavin turns one photo into a five-minute film and lands the category’s known failure as the payoff: “the thing that really blew my mind was the character consistency across the whole five minutes. Everything from wide shots to close-ups, it never broke. Something that most models struggle with” (0:49). Three creators, three weeks, three different jobs: define the word, credit the brand for it, prove the promise. None of them read the same script.

The second wave is repetition. On August 17 alone, five TikTok accounts whose recent feeds are almost entirely OpenArt each published a video that says the word. ugcnatee: “It’s gonna vibe direct you a professional walkthrough of the course up to five minutes long” (0:18). eddielikesai: “It’ll vibe direct you a walkthrough video like this one” (0:22). Same verb, same feature, same “up to five minutes.”

Did the word take? We can’t see search volume or comment sections, so adoption is unmeasured. What is measured: the most-viewed post that says it is a YouTube tutorial by Tech Vorld at 98K, exactly 1.0× that channel’s normal. The word travelled. The views did not travel with it.

On YouTube, the winners show everything

OpenArt’s ten most-viewed postsWhat the top of the sample has in common.
9 of 10are YouTube videos
9 of 10carry a link in the description
9 of 10put the product interface on screen, not hidden
The one exception is a silent cinematic TikTok with a comment-keyword ask. “Hide the product” is not what the winners here do.

The biggest one, Shark Numbers at 1.36 million views, has a structure you can copy in four beats. Play it, then read the beats against the frames.

Shark Numbers holding a shark that used to be a toy
Can AI VFX Handle Real Video Editing? · Shark Numbers1,358,340 views · 1.94M subscribers · 1.36× the channel’s median · affiliate tracking link in the description · like ratio 4.2% against a brand median of 2.9%
Frame from the Shark Numbers reviewBEAT 1 OF 4

Now the same top three, measured against each creator’s own baseline instead of raw views. Raw views make them look like stronger and weaker versions of the same thing. The multiple shows three different purchases.

Paid posts in this sample land at a median of 1.08× the creator’s baseline, which is what buying reach looks like: you get the channel’s normal audience, not a hit. Affiliate posts sit at 2.05×, and 13 of the 63 with a baseline beat 10×, so that layer is where the brand’s breakouts actually come from. The contest films sit at the other end: ten entries, median 196 views, zero calls to action, product credited only in the description. OpenArt is not buying reach there. It is buying a portfolio it can show.

The cluster that got the word but not the views

Ten TikTok accounts in this sample devote all of their last 30 posts to OpenArt. Two carry a program tag; the other eight have no public relationship on record. We do not claim OpenArt runs them, and nothing public says so. What we can see is what they post.

On August 4, ksdealz.ugc published a script: screenshot a local golf course, feed it to OpenArt Director, “vibe direct” a flyover video, sell it to the course for $1,000. It did 77,500 views on an account with 4,345 followers. Then the script was copied.

The copies share more than the pitch. Savanna Vaden and MadeByKbb open on the same stock clip. eddielikesai and rikki open with the same duet layout and the same first line over a Porsche 911 challenge. Three of them close on the same sentence, in slightly different words: eddielikesai, “After doing this for a month, I have no clue why nobody else is talking about this” (0:33); rikki, “Doing it last month, I don’t know why no one is talking about it” (0:26).

Five copies of a 77K script produced 2,197 views between them. Program-tagged posts in this sample have a median of 493 views. This is the floor of the seeding machine: the language spreads on schedule, the views do not. Every account is a separate lottery ticket, and the ticket that hit was the first one.

Our library is a winners’ sample. We cannot see the brand-hiding films nobody watched, or the tutorials that never ranked. Judge a play by the batch, not the post.

Three lines for your next brief

If your product makes something people would watch on its own, here is what OpenArt’s sample supports.

01
Split creators by what the viewer wants, not by platform.

One brief per line: learn, compare, watch, earn, fix the known failure. Never ask one creator to do two lines in one video.

02
For reviews, put the real interface and the real account on screen.

Workspace name, credit balance, price per generation, the creator’s own footage as the test. Then measure the post against that channel’s baseline. 1.0× is what you paid for; 10× is luck you should study, not budget for.

03
If you want a word to stick, hand it to three people who will explain it differently before anyone recites it.

Definition, attribution, proof. The recitation layer comes last, and expect it to do hundreds of views, not thousands.

If you are starting from zero: the four tiers OpenArt runs, in the order to build them

OpenArt asks Creative Partners for no posting quota: “There are no strict rules — we're here to support your wildest content ideas.” OpenArt offers them “Free credits to OpenArt,” “Pre-launch access of new features,” “Spotlight on social media and websites,” “Priority paid collaboration opportunities,” “Access to exclusive community events,” and “Pre-approval for our Affiliate Program.” The program posts in this sample reached a 493-view median, while the affiliate tier produced the breakouts above. Run the program tier as your intake for affiliates, and judge it by how many partners convert into tracked affiliates, not by the views its posts get.

1
AFFILIATE Tracking links with a personal code

64 creators here, links like openart.ai/…?ref=name in the description. Costs commission only. 13 of the 63 measurable affiliate posts beat 10× the creator’s baseline: this is where the hits came from. Build this first.

2
PROGRAM A named partner program with a credit line

19 creators tagged #openartcpp or “Creative Partner Program.” Median 493 views. It buys a steady stream of tutorials and a portfolio, not reach. Judge it by how many partners convert into tracked affiliates, never by views.

3
CONTEST Film competitions with a partner studio

9 creators, entries to the Wonder Studios AI Director Challenge and MiniMax Story Weeks. Median 196 views, zero calls to action. You get finished films you can show; you do not get distribution.

4
PAID Disclosed placements on channels you choose

13 creators with #OpenArtPartner or “sponsored by.” Median 1.08× the channel’s baseline. Buy this last, and only for a channel whose existing audience is the one you want.

Brand-owned program pageArchived relationship receipt
OpenArt Creative Partner Program page with its description and an Apply Now button
SOURCE · https://openart.ai/programs/cpp · CAPTURED · 2026-09-02 · WAYBACK ↗ · LOWER-PAGE COPY QUOTED ABOVE WAS VERIFIED AT CAPTURE AND SITS OUTSIDE THIS CROP

The brief for the format that won, ready to paste

BRIEF · PRODUCT REVIEW · YOUTUBE · 8–10 MIN
  1. Open on the finished result. No intro, no logo. Under two seconds.
  2. State the problem in the words your trade uses, not in ours.
  3. Say the doubt your viewers already have, out loud, before you test.
  4. Test it on your own footage, in the real interface, with your own account and credit balance in frame.
  5. Tracking link in the description. Disclose the relationship in the first minute.
  6. We will read this post against your channel’s median views, not raw views. 1× is fine. We are buying your audience, not a lottery ticket.
Built from Shark Numbers’ four beats (1.36M views, 1.36× baseline, 4.2% like ratio). Swap in your own product’s known failure at step 2.

If your product’s output can’t be filmed, none of this transfers. Your problem is what to show instead, and that is a different piece.

The word is the cheapest thing a brand can distribute. It is also the only thing in this sample that spread on its own.

Sources & methodology

We studied 250 public creator posts connected to OpenArt. They show six distinct creator lines, a coordinated vocabulary, and a dedicated-account cluster. Public posts cannot tell us how many customers OpenArt acquired from any of it.

The operation · all-time tracked samplepublic creator posts, not OpenArt’s books
250posts audited
227creators traced
6scripts that repeat
8posts over 100K views
PAID 14AFFILIATE 73PROGRAM 19CONTEST 10UNKNOWN 133unknown is never treated as organic
YouTube 153TikTok 57Instagram 40
METHOD · 250 public posts from 227 creators, audited August 2026. Multiples are a post’s views divided by that creator’s median views over their recent feed. Relationship labels require a public disclosure, a sponsor statement, a program credit, or a trackable affiliate link; “unknown” means no receipt was found, not that none exists. “Vibe directing” counts are full-text matches in transcripts. Instagram hides view counts, so Instagram anchors are cited without views. Timestamps come from full transcripts; first-frame descriptions come from stored frames. This is observation, not a controlled test.