Creators say Wan 3.0 can remake a 23-million-view TikTok for $0.20, and that matters if you make AI ads
Three independent creator posts show Wan 3.0 producing UGC-style ads, 15-second multi-shot scenes and short dramas at very low cost. Here is what to test first, and what nobody has confirmed yet.
One creator says he took a TikTok that pulled 23 million views and remade it with Wan 3.0 for $0.20. Twenty cents. If that holds up in your hands, the economics of the AI UGC ad just moved again.
The claim comes from @paul_popov, who posted the side by side with a blunt line: "Don't sleep on AI UGC." He is not alone in pointing at Wan 3.0 this week. @iamahmedfaraz66 shared a full prompt for a 15-second, 16:9 piece with a single character held across every shot. And @miiarosee7 posted a short called "The Last Customer", a boy, a birthday cake and a baker, made with Wan 3.0 on @budgetpixel. Three different people, three different use cases, one model.
What the posts actually show
The $0.20 figure is the headline, and it is the weakest piece of evidence. It is one creator's own claim, per @paul_popov, and he has not said how many attempts it took, what resolution he rendered at, or whether the number includes failed generations. Anecdotes are not benchmarks. Treat it as a reason to run your own test, not as a price list.
The prompt from @iamahmedfaraz66 is more useful to a filmmaker. It asks for a 15-second ultra-realistic 16:9 video, locks the character's face, hair, outfit, scarf and body proportions, and then breaks the clip into timed shots, starting with 0 to 3 seconds on a quiet autumn street. That is the structure working filmmakers already use with other current models. It suggests Wan 3.0 accepts timecoded multi-shot direction and an identity lock in a single prompt. Whether it keeps that identity steady across all five shots is something the post does not prove, and a single clip cannot.
@miiarosee7 shows the other end of the range. A small emotional story, with a child, a baker and a cake, is the sort of scene that punishes uncanny hands and stiff faces. It is a good sign that it was attempted. It is not a stress test.
How it sits against the alternatives
The obvious comparison is Seedance 2.5, which dominates the same timeline with 30-second continuous generations. Wan 3.0 is being shown at 15 seconds in these posts, so if you need long single takes, the evidence here does not put it ahead. Where it may win is price. If the $0.20 figure is even roughly representative, it undercuts the per-second rates quoted for the hosted alternatives, and for short-form ad work that is the number that decides which model gets the job.
That is also why the UGC angle makes sense. A 15-second talking-head ad does not need a 30-second take. It needs a believable face, steady hands, and a low cost per attempt so you can throw away nine versions and keep one.
What has not been said
There is no official announcement in any of these posts. We have no release date from the model's makers, no published pricing, no licence terms and no statement on whether the weights are open. One post names a host, @budgetpixel, but nothing says that is the only place to run it or what it charges. Nobody has shown the clip lengths beyond 15 seconds, the maximum resolution, or how audio is handled.
And a remake of a viral video proves less than it sounds. Copying a format that already worked is easier than inventing one, and the 23 million views belong to the original, not to the remake.
What to test first
1. Cost per keeper. Generate ten versions of the same 15-second ad and divide the total spend by the number you would actually publish. That is your real price, not $0.20.
2. Identity across shots. Copy the structure from @iamahmedfaraz66: a locked character, timecoded shots, and a reference image. Check the face at shot four, not shot one.
3. Emotional close-ups. Run a scene like "The Last Customer" with a child and hands holding an object. Hands and food are where cheap models break.
4. Rights. If you are remaking an existing viral clip for a client, clear the format and any recognisable talent first.
If the numbers hold, Wan 3.0 becomes your test-bench model for volume work, and you keep the pricier options for the hero shots.
Sources: @paul_popov (the $0.20 TikTok remake), @iamahmedfaraz66 (the 15-second Wan 3.0 prompt), @miiarosee7 ("The Last Customer" short on budgetpixel).