How the Coyote vs Acme audience chose hope over burnout, TikTok marketing analysis with anex.sh

The movie was a hit in Latin America

The film opened in cinemas at the end of August 2026. We collected 2,264 TikTok posts about it and had open models read every one. Accounts in Latin America posted 878 of them (38.8%) and drew 52.7% of all 274 million views. Accounts in the United States posted a few more, 898 (39.7%), and drew 31.4%.

Here is the median number of views per post, by the country the account is registered in:

Grouped by the language of the caption instead of the account's region, posts with a Spanish caption are 22.3% of all posts and hold 33.6% of the views, with a median of 14,600 views per post. Posts with an English caption have a median of 4,508.

The most viewed post in the whole collection is a Mexican recap of the scene where the Coyote finally reaches the Road Runner. It has 7.9 million views and came from an account with 4,461 followers.

Marketing played a part. Mexico's distributor, Zima Entertainment, tagged Profeco, the Mexican consumer-protection agency, in a mock complaint against ACME. 20 posts mention Profeco. Creators covered the stunt as news worth sharing: one post calls it the most original ad campaign Mexico has seen (360,900 views) and reports that the newspaper Milenio gave its front page to the lawsuit as if it were a real case.

Fans also treated ticket sales as a regional effort. Of the 145 posts about the box office, 44.8% have Spanish captions, twice the share across the collection.

One scene did the work

In the classic cartoons the Coyote never catches the Road Runner. In the film he finally touches him. Fans count the wait as 77 years, from the first short in 1949. Fans also dug up where his silence comes from: in the 2000 Looney Tunes short Little Go Beep, his father tells the young Coyote he may not speak until he catches a roadrunner.

That one beat is the emotional centre of the film on TikTok.

• 71 captions name the touch or the catch, with 16.7 million views between them.
• The father and the vow of silence appear in 59 captions. Those posts have a median of 57,900 views, nearly twelve times the median for the whole collection.
• The top three posts that name the scene come from Mexico, Venezuela and the United States, in two languages.
• The comments turn it into a shared joke and a shared grief. On the most viewed post, the top comment (220,000 likes) points out that he does it all to be able to talk to his father again.

The music follows the scene. ABBA's "The Winner Takes It All" is on 18 posts, 12 of them in Spanish, with a median of 77,250 views.

Posts with an emotional tone do better than hype. A model sorted every post by its tone. The 158 posts it read as tender have a median of 21,750 views. The 912 posts it read as hype have a median of 4,267.

Two examples:

• A Venezuelan picture carousel retelling the touch with the father's instruction, 1.8 million views.
• A US video asking why the Coyote cried after catching the Road Runner, 1.5 million views.

The Coyote became a never-give-up icon

Only a few fans used the Coyote as a symbol of adult life going wrong: burnout, debt, a bad job.

A model read every post and noted when the creator uses the Coyote or ACME to stand for something in real life. It found 298 such posts out of 2,264. Grouped:

Only 2 posts name burnout outright. One of them has 209,200 views.

What spread instead is a plain motivational line: if he never gave up, why should we? In Spanish that comment has 130,000 likes on the most viewed post, and it is the caption of a Mexican edit with 1.1 million views. The English version has 29,000 likes on another post.

The comments agree. Of 13,954 collected comments, 68 carry a never-give-up line and 35 treat the Coyote as a burnout or "literally me" figure.

The reading travels furthest in Spanish. Persistence is about as common in English captions (15.9%) as in Spanish ones (14.8%). The Spanish posts reach much further: a median of 97,300 views, against 13,350 in English.

Fans drew a parallel between the real studio fight and the movie's plot

The film has a real backstory. Warner Bros. shelved it, and in March 2025 another distributor, Ketchup Entertainment, rescued it. The film's plot is the Coyote suing ACME.

Fans told the rescue as a real-life Coyote versus ACME, and treated buying a ticket as taking his side.

• 337 posts (14.9%) cover the studio backstory. Most are news accounts with a low median of 3,189 views.
• The 43 posts that make the studio the real-life stand-in for ACME do much better: a median of 19,500 views.
• Commenters name Warner as the villain and ask others to see the film in cinemas so ACME does not win.
• 332 captions ask people to see the film in a cinema. 221 of them come from accounts under 10,000 followers.

The posts before release come in four small waves: the shelving news in February 2024, the Animation Guild's call to release the film in June 2024, the rescue in March 2025 and the trailer in April 2026.

• The Animation Guild's call to release the film, 2.6 million views.
• A Spanish-language celebration of the rescue, 912,000 views.

The movie trended in two distinct waves

Release week belonged to news, reviews and promotion. Two weeks later, the posts were clip edits and fan art from small accounts.

The busiest day in the collection is 13 September, with 264 posts. That is the peak of the second wave, more than two weeks after release.

Fan-made posts get the most likes per view: 16.6% for fan art and 10.6% for clip edits, against 4.5% for news and 4.3% for official promotion.

Small accounts made most of the hits. The top 10% of posts by views is 227 posts. 128 of them come from accounts under 10,000 followers. 18 come from verified accounts.

Part of the fan-art rise is a trend of drawing the Road Runner as a human. It appears from 6 September on, led by one such drawing with 1.7 million views. Another big fan post is a tribute edit from Greece with 2.3 million views from an account of 2,152 followers.

Credit to the marketing team: meep meep or beep beep

The film's marketing team hit the nail on the head. At release, the official account asked one small question: Team Beep or Team Meep? It is an old argument about how the Road Runner's call is spelled.

They planted the seed and it took off. The argument comes up in the comment threads of 205 posts, more than any other scene or joke we measured. The Coyote crying is next, at 162. Only 58 captions raise it, so the audience carried it on its own.

All 13 running jokes and scenes we counted, by the number of posts whose comments mention them:

• The official post, 777,400 views. Its top two comments are the argument itself.
• An explainer that settles it as meep meep, 315,200 views. It shows how the Road Runner theme song was built from Paul Julian's single vocal track.

FPS matters for video analysis

A model does not watch a video the way a person does. It looks at still frames taken at a set rate, measured in frames per second (fps). Here is how the big hosted APIs handle it, according to their documentation in September 2026:

• anex.sh sets the rate per job with the video_fps field. On Qwen 3.8 27B it defaults to 1 fps and goes up to 20 fps.
• Google Gemini, on its API and on Vertex AI, samples 1 fps by default and lets you set a rate of up to 24 fps per video.
• Amazon Nova Lite and Pro on Bedrock sample 1 fps for videos up to 16 minutes and a fixed 960 frames for longer ones, with no documented setting to change it.
• Alibaba's Qwen API samples 2 fps by default, adjustable from 0.1 to 10 fps.
• OpenAI and Anthropic do not take video input: you extract frames yourself and send them as images.

We read the same video clips at 1 fps and again at 5 fps, which costs about five times as many tokens per clip. The higher rate catches text that is on screen for only a moment:

• An edit with 1.6 million views shows a 3D title calling the pair legendary lovers, late in its second 30-second segment. Only the 5 fps run reported it.
• A smaller edit shows a second sign card asking about fate for about a second. The 1 fps run skipped it.

How we did it

After we downloaded the most viewed and liked TikTok posts using Bright Data's TikTok keyword search, we processed them with anex.sh in batches.

Models. All reading ran as batch jobs on anex.sh, each job one input file.

Text, every post: Qwen 3.8 27B

Two passes over all 2,264 captions and comments. The first described themes and any relationship reading. The second asked what the creator is using the film for: format, emotional tone, intended audience and any real-life stand-in.

Video and images, 288 posts: Qwen 3.8 27B

Videos were cut into 30-second segments and read at 1 fps and again at 5 fps, using the job-level video_fps field. Carousels were read two images at a time. A separate pass labelled the visual format of these posts.

Audio, video posts in the sample: Gemma 4 12B

Speech, music mood and short intelligible phrases, in 30-second segments.

Combining: Qwen 3.8 27B

One text-only request per post holding the caption reading, every visual and audio reading and a note of any missing media, once with the 1 fps readings and once with the 5 fps ones.

Cost.

The 5 fps rate is what drives the video cost. A 30-second segment is about 3,300 input tokens at 1 fps and about 15,900 at 5 fps.

Time. The first round of 24 jobs, from the first submission to the last downloaded result, took about 3 hours 45 minutes on 14 September, including jobs we cancelled and reissued. Each combining job took about 21 minutes. The second text pass and the visual format pass ran afterwards, as two more jobs.

Limitations

• This is not all of TikTok. The posts come from a keyword search, so we might have missed some posts altogether.
• Engagement is one snapshot. Posts from the last few days before 14 September had little time to gather views, so the size of the second wave is uncertain. Nothing here measures growth over time.
• Region is the account's region, not where the viewers are. The Latin American release was a day before the US one.